wip /api/chat

2025-12-26 00:58:12 -05:00 · 2023-10-01 14:54:17 -07:00
82 changed files with 789 additions and 2359 deletions
--- a/6
+++ b/6
@@ -1,16 +1,16 @@
 FROM nvidia/cuda:11.8.0-devel-ubuntu22.04

 ARG TARGETARCH
+ARG VERSION=0.0.0
 ARG GOFLAGS="'-ldflags=-w -s'"

 WORKDIR /go/src/github.com/jmorganca/ollama
 RUN apt-get update && apt-get install -y git build-essential cmake
-ADD https://dl.google.com/go/go1.21.3.linux-$TARGETARCH.tar.gz /tmp/go1.21.3.tar.gz
-RUN mkdir -p /usr/local && tar xz -C /usr/local </tmp/go1.21.3.tar.gz
+ADD https://dl.google.com/go/go1.21.1.linux-$TARGETARCH.tar.gz /tmp/go1.21.1.tar.gz
+RUN mkdir -p /usr/local && tar xz -C /usr/local </tmp/go1.21.1.tar.gz

 COPY . .
 ENV GOARCH=$TARGETARCH
-ENV GOFLAGS=$GOFLAGS
 RUN /usr/local/go/bin/go generate ./... \
    && /usr/local/go/bin/go build .

--- a/Dockerfile.build
+++ b/Dockerfile.build
@@ -1,5 +1,6 @@
+
 # centos7 amd64 dependencies
-FROM --platform=linux/amd64 nvidia/cuda:11.3.1-devel-centos7 AS base-amd64
+FROM --platform=linux/amd64 nvidia/cuda:11.8.0-devel-centos7 AS base-amd64
 RUN yum install -y https://repo.ius.io/ius-release-el7.rpm centos-release-scl && \
    yum update -y && \
    yum install -y devtoolset-10-gcc devtoolset-10-gcc-c++ git236 wget
@@ -7,25 +8,25 @@ RUN wget "https://github.com/Kitware/CMake/releases/download/v3.27.6/cmake-3.27.
 ENV PATH /opt/rh/devtoolset-10/root/usr/bin:$PATH

 # centos8 arm64 dependencies
-FROM --platform=linux/arm64 nvidia/cuda-arm64:11.3.1-devel-centos8 AS base-arm64
+FROM --platform=linux/arm64 nvidia/cuda:11.4.3-devel-centos8 AS base-arm64
 RUN sed -i -e 's/mirrorlist/#mirrorlist/g' -e 's|#baseurl=http://mirror.centos.org|baseurl=http://vault.centos.org|g' /etc/yum.repos.d/CentOS-*
 RUN yum install -y git cmake

 FROM base-${TARGETARCH}
 ARG TARGETARCH
-ARG GOFLAGS="'-ldflags -w -s'"

 # install go
-ADD https://dl.google.com/go/go1.21.3.linux-$TARGETARCH.tar.gz /tmp/go1.21.3.tar.gz
-RUN mkdir -p /usr/local && tar xz -C /usr/local </tmp/go1.21.3.tar.gz
+ADD https://dl.google.com/go/go1.21.1.linux-$TARGETARCH.tar.gz /tmp/go1.21.1.tar.gz
+RUN mkdir -p /usr/local && tar xz -C /usr/local </tmp/go1.21.1.tar.gz

 # build the final binary
 WORKDIR /go/src/github.com/jmorganca/ollama
 COPY . .
-
 ENV GOOS=linux
 ENV GOARCH=$TARGETARCH
-ENV GOFLAGS=$GOFLAGS
+
+ARG VERSION=0.0.0
+ARG GOFLAGS="'-ldflags -w -s'"

 RUN /usr/local/go/bin/go generate ./... && \
    /usr/local/go/bin/go build .
--- a/README.md
+++ b/README.md
@@ -13,7 +13,7 @@ Get up and running with large language models locally.

 ### macOS

-[Download](https://ollama.ai/download/Ollama-darwin.zip)
+[Download](https://ollama.ai/download/Ollama-darwin.zip) 

 ### Linux & WSL2

@@ -23,7 +23,7 @@ curl https://ollama.ai/install.sh | sh

 [Manual install instructions](https://github.com/jmorganca/ollama/blob/main/docs/linux.md)

-### Windows
+### Windows 

 coming soon

@@ -37,7 +37,7 @@ ollama run llama2

 ## Model library

-Ollama supports a list of open-source models available on [ollama.ai/library](https://ollama.ai/library 'ollama model library')
+Ollama supports a list of open-source models available on [ollama.ai/library](https://ollama.ai/library "ollama model library")

 Here are some example open-source models that can be downloaded:

@@ -66,13 +66,13 @@ Ollama supports importing GGUF and GGML file formats in the Modelfile. This mean
   FROM ./vicuna-33b.Q4_0.gguf
   ```

-2. Create the model in Ollama
+3. Create the model in Ollama

   ```
   ollama create name -f path_to_modelfile
   ```

-3. Run the model
+5. Run the model

   ```
   ollama run name
@@ -109,7 +109,7 @@ ollama run mario
 Hello! It's your friend Mario.
 ```

-For more examples, see the [examples](examples) directory. For more information on working with a Modelfile, see the [Modelfile](docs/modelfile.md) documentation.
+For more examples, see the [examples](./examples) directory. For more information on working with a Modelfile, see the [Modelfile](./docs/modelfile.md) documentation.

 ## CLI Reference

@@ -195,7 +195,7 @@ Finally, in a separate shell, run a model:

 ## REST API

-> See the [API documentation](docs/api.md) for all endpoints.
+> See the [API documentation](./docs/api.md) for all endpoints.

 Ollama has an API for running and managing models. For example to generate text from a model:

@@ -217,7 +217,6 @@ curl -X POST http://localhost:11434/api/generate -d '{
 - [Dagger Chatbot](https://github.com/samalba/dagger-chatbot)
 - [LiteLLM](https://github.com/BerriAI/litellm)
 - [Discord AI Bot](https://github.com/mekb-turtle/discord-ai-bot)
- [Chatbot UI](https://github.com/ivanfioravanti/chatbot-ollama)
 - [HTML UI](https://github.com/rtcfirefly/ollama-ui)
 - [Typescript UI](https://github.com/ollama-interface/Ollama-Gui?tab=readme-ov-file)
 - [Dumbar](https://github.com/JerrySievert/Dumbar)
--- a/api/client.go
+++ b/api/client.go
@@ -7,7 +7,6 @@ import (
 	"encoding/json"
 	"fmt"
 	"io"
-	"net"
 	"net/http"
 	"net/url"
 	"os"
@@ -19,11 +18,14 @@ import (

 const DefaultHost = "127.0.0.1:11434"

-var envHost = os.Getenv("OLLAMA_HOST")
+var (
+	envHost = os.Getenv("OLLAMA_HOST")
+)

 type Client struct {
-	base *url.URL
-	http http.Client
+	Base    url.URL
+	HTTP    http.Client
+	Headers http.Header
 }

 func checkError(resp *http.Response, body []byte) error {
@@ -42,44 +44,34 @@ func checkError(resp *http.Response, body []byte) error {
 	return apiError
 }

-func ClientFromEnvironment() (*Client, error) {
-	scheme, hostport, ok := strings.Cut(os.Getenv("OLLAMA_HOST"), "://")
-	if !ok {
-		scheme, hostport = "http", os.Getenv("OLLAMA_HOST")
+// Host returns the default host to use for the client. It is determined in the following order:
+// 1. The OLLAMA_HOST environment variable
+// 2. The default host (localhost:11434)
+func Host() string {
+	if envHost != "" {
+		return envHost
+	}
+	return DefaultHost
+}
+
+// FromEnv creates a new client using Host() as the host. An error is returns
+// if the host is invalid.
+func FromEnv() (*Client, error) {
+	h := Host()
+	if !strings.HasPrefix(h, "http://") && !strings.HasPrefix(h, "https://") {
+		h = "http://" + h
 	}

-	host, port, err := net.SplitHostPort(hostport)
+	u, err := url.Parse(h)
 	if err != nil {
-		host, port = "127.0.0.1", "11434"
-		if ip := net.ParseIP(strings.Trim(os.Getenv("OLLAMA_HOST"), "[]")); ip != nil {
-			host = ip.String()
-		}
+		return nil, fmt.Errorf("could not parse host: %w", err)
 	}

-	client := Client{
-		base: &url.URL{
-			Scheme: scheme,
-			Host:   net.JoinHostPort(host, port),
-		},
+	if u.Port() == "" {
+		u.Host += ":11434"
 	}

-	mockRequest, err := http.NewRequest("HEAD", client.base.String(), nil)
-	if err != nil {
-		return nil, err
-	}
-
-	proxyURL, err := http.ProxyFromEnvironment(mockRequest)
-	if err != nil {
-		return nil, err
-	}
-
-	client.http = http.Client{
-		Transport: &http.Transport{
-			Proxy: http.ProxyURL(proxyURL),
-		},
-	}
-
-	return &client, nil
+	return &Client{Base: *u, HTTP: http.Client{}}, nil
 }

 func (c *Client) do(ctx context.Context, method, path string, reqData, respData any) error {
@@ -94,7 +86,7 @@ func (c *Client) do(ctx context.Context, method, path string, reqData, respData
 		reqBody = bytes.NewReader(data)
 	}

-	requestURL := c.base.JoinPath(path)
+	requestURL := c.Base.JoinPath(path)
 	request, err := http.NewRequestWithContext(ctx, method, requestURL.String(), reqBody)
 	if err != nil {
 		return err
@@ -104,7 +96,11 @@ func (c *Client) do(ctx context.Context, method, path string, reqData, respData
 	request.Header.Set("Accept", "application/json")
 	request.Header.Set("User-Agent", fmt.Sprintf("ollama/%s (%s %s) Go/%s", version.Version, runtime.GOARCH, runtime.GOOS, runtime.Version()))

-	respObj, err := c.http.Do(request)
+	for k, v := range c.Headers {
+		request.Header[k] = v
+	}
+
+	respObj, err := c.HTTP.Do(request)
 	if err != nil {
 		return err
 	}
@@ -127,8 +123,6 @@ func (c *Client) do(ctx context.Context, method, path string, reqData, respData
 	return nil
 }

-const maxBufferSize = 512 * 1000 // 512KB
-
 func (c *Client) stream(ctx context.Context, method, path string, data any, fn func([]byte) error) error {
 	var buf *bytes.Buffer
 	if data != nil {
@@ -140,26 +134,23 @@ func (c *Client) stream(ctx context.Context, method, path string, data any, fn f
 		buf = bytes.NewBuffer(bts)
 	}

-	requestURL := c.base.JoinPath(path)
+	requestURL := c.Base.JoinPath(path)
 	request, err := http.NewRequestWithContext(ctx, method, requestURL.String(), buf)
 	if err != nil {
 		return err
 	}

 	request.Header.Set("Content-Type", "application/json")
-	request.Header.Set("Accept", "application/x-ndjson")
+	request.Header.Set("Accept", "application/json")
 	request.Header.Set("User-Agent", fmt.Sprintf("ollama/%s (%s %s) Go/%s", version.Version, runtime.GOARCH, runtime.GOOS, runtime.Version()))

-	response, err := c.http.Do(request)
+	response, err := http.DefaultClient.Do(request)
 	if err != nil {
 		return err
 	}
 	defer response.Body.Close()

 	scanner := bufio.NewScanner(response.Body)
-	// increase the buffer size to avoid running out of space
-	scanBuf := make([]byte, 0, maxBufferSize)
-	scanner.Buffer(scanBuf, maxBufferSize)
 	for scanner.Scan() {
 		var errorResponse struct {
 			Error string `json:"error,omitempty"`
--- a/api/types.go
+++ b/api/types.go
@@ -31,13 +31,28 @@ func (e StatusError) Error() string {
 	}
 }

+// /api/chat
+type Message struct {
+	Role    string `json:"role"`
+	Content string `json:"content"`
+}
+
+type ChatRequest struct {
+	Model    string    `json:"model"`
+	Messages []Message `json:"messages"`
+}
+
+type ChatResponse struct {
+	CreatedAt time.Time `json:"created_at"`
+	Message   Message   `json:"message"`
+}
+
 type GenerateRequest struct {
 	Model    string `json:"model"`
 	Prompt   string `json:"prompt"`
 	System   string `json:"system"`
 	Template string `json:"template"`
 	Context  []int  `json:"context,omitempty"`
-	Stream   *bool  `json:"stream,omitempty"`

 	Options map[string]interface{} `json:"options"`
 }
@@ -54,9 +69,8 @@ type EmbeddingResponse struct {
 }

 type CreateRequest struct {
-	Name   string `json:"name"`
-	Path   string `json:"path"`
-	Stream *bool  `json:"stream,omitempty"`
+	Name string `json:"name"`
+	Path string `json:"path"`
 }

 type DeleteRequest struct {
@@ -83,9 +97,6 @@ type CopyRequest struct {
 type PullRequest struct {
 	Name     string `json:"name"`
 	Insecure bool   `json:"insecure,omitempty"`
-	Username string `json:"username"`
-	Password string `json:"password"`
-	Stream   *bool  `json:"stream,omitempty"`
 }

 type ProgressResponse struct {
@@ -98,9 +109,6 @@ type ProgressResponse struct {
 type PushRequest struct {
 	Name     string `json:"name"`
 	Insecure bool   `json:"insecure,omitempty"`
-	Username string `json:"username"`
-	Password string `json:"password"`
-	Stream   *bool  `json:"stream,omitempty"`
 }

 type ListResponse struct {
@@ -121,7 +129,7 @@ type TokenResponse struct {
 type GenerateResponse struct {
 	Model     string    `json:"model"`
 	CreatedAt time.Time `json:"created_at"`
-	Response  string    `json:"response"`
+	Response  string    `json:"response,omitempty"`

 	Done    bool  `json:"done"`
 	Context []int `json:"context,omitempty"`
@@ -205,8 +213,6 @@ type Options struct {
 	NumThread int `json:"num_thread,omitempty"`
 }

-var ErrInvalidOpts = fmt.Errorf("invalid options")
-
 func (opts *Options) FromMap(m map[string]interface{}) error {
 	valueOpts := reflect.ValueOf(opts).Elem() // names of the fields in the options struct
 	typeOpts := reflect.TypeOf(opts).Elem()   // types of the fields in the options struct
@@ -220,7 +226,6 @@ func (opts *Options) FromMap(m map[string]interface{}) error {
 		}
 	}

-	invalidOpts := []string{}
 	for key, val := range m {
 		if opt, ok := jsonOpts[key]; ok {
 			field := valueOpts.FieldByName(opt.Name)
@@ -238,12 +243,12 @@ func (opts *Options) FromMap(m map[string]interface{}) error {
 						// when JSON unmarshals numbers, it uses float64, not int
 						field.SetInt(int64(t))
 					default:
-						log.Printf("could not convert model parameter %v of type %T to int, skipped", key, val)
+						log.Printf("could not convert model parameter %v to int, skipped", key)
 					}
 				case reflect.Bool:
 					val, ok := val.(bool)
 					if !ok {
-						log.Printf("could not convert model parameter %v of type %T to bool, skipped", key, val)
+						log.Printf("could not convert model parameter %v to bool, skipped", key)
 						continue
 					}
 					field.SetBool(val)
@@ -251,14 +256,14 @@ func (opts *Options) FromMap(m map[string]interface{}) error {
 					// JSON unmarshals to float64
 					val, ok := val.(float64)
 					if !ok {
-						log.Printf("could not convert model parameter %v of type %T to float32, skipped", key, val)
+						log.Printf("could not convert model parameter %v to float32, skipped", key)
 						continue
 					}
 					field.SetFloat(val)
 				case reflect.String:
 					val, ok := val.(string)
 					if !ok {
-						log.Printf("could not convert model parameter %v of type %T to string, skipped", key, val)
+						log.Printf("could not convert model parameter %v to string, skipped", key)
 						continue
 					}
 					field.SetString(val)
@@ -266,7 +271,7 @@ func (opts *Options) FromMap(m map[string]interface{}) error {
 					// JSON unmarshals to []interface{}, not []string
 					val, ok := val.([]interface{})
 					if !ok {
-						log.Printf("could not convert model parameter %v of type %T to slice, skipped", key, val)
+						log.Printf("could not convert model parameter %v to slice, skipped", key)
 						continue
 					}
 					// convert []interface{} to []string
@@ -274,7 +279,7 @@ func (opts *Options) FromMap(m map[string]interface{}) error {
 					for i, item := range val {
 						str, ok := item.(string)
 						if !ok {
-							log.Printf("could not convert model parameter %v of type %T to slice of strings, skipped", key, item)
+							log.Printf("could not convert model parameter %v to slice of strings, skipped", key)
 							continue
 						}
 						slice[i] = str
@@ -284,51 +289,45 @@ func (opts *Options) FromMap(m map[string]interface{}) error {
 					return fmt.Errorf("unknown type loading config params: %v", field.Kind())
 				}
 			}
-		} else {
-			invalidOpts = append(invalidOpts, key)
 		}
 	}
-
-	if len(invalidOpts) > 0 {
-		return fmt.Errorf("%w: %v", ErrInvalidOpts, strings.Join(invalidOpts, ", "))
-	}
 	return nil
 }

 func DefaultOptions() Options {
 	return Options{
-		// options set on request to runner
-		NumPredict:       -1,
-		NumKeep:          -1,
+		Seed: -1,
+
+		UseNUMA: false,
+
+		NumCtx:             2048,
+		NumKeep:            -1,
+		NumBatch:           512,
+		NumGPU:             -1, // -1 here indicates that NumGPU should be set dynamically
+		NumGQA:             1,
+		LowVRAM:            false,
+		F16KV:              true,
+		UseMMap:            true,
+		UseMLock:           false,
+		RopeFrequencyBase:  10000.0,
+		RopeFrequencyScale: 1.0,
+		EmbeddingOnly:      true,
+
+		RepeatLastN:      64,
+		RepeatPenalty:    1.1,
+		FrequencyPenalty: 0.0,
+		PresencePenalty:  0.0,
 		Temperature:      0.8,
 		TopK:             40,
 		TopP:             0.9,
 		TFSZ:             1.0,
 		TypicalP:         1.0,
-		RepeatLastN:      64,
-		RepeatPenalty:    1.1,
-		PresencePenalty:  0.0,
-		FrequencyPenalty: 0.0,
 		Mirostat:         0,
 		MirostatTau:      5.0,
 		MirostatEta:      0.1,
 		PenalizeNewline:  true,
-		Seed:             -1,

-		// options set when the model is loaded
-		NumCtx:             2048,
-		RopeFrequencyBase:  10000.0,
-		RopeFrequencyScale: 1.0,
-		NumBatch:           512,
-		NumGPU:             -1, // -1 here indicates that NumGPU should be set dynamically
-		NumGQA:             1,
-		NumThread:          0, // let the runtime decide
-		LowVRAM:            false,
-		F16KV:              true,
-		UseMLock:           false,
-		UseMMap:            true,
-		UseNUMA:            false,
-		EmbeddingOnly:      true,
+		NumThread: 0, // let the runtime decide
 	}
 }

--- a/cmd/cmd.go
+++ b/cmd/cmd.go
@@ -61,7 +61,7 @@ func CreateHandler(cmd *cobra.Command, args []string) error {
 		return err
 	}

-	client, err := api.ClientFromEnvironment()
+	client, err := api.FromEnv()
 	if err != nil {
 		return err
 	}
@@ -119,7 +119,7 @@ func CreateHandler(cmd *cobra.Command, args []string) error {
 }

 func RunHandler(cmd *cobra.Command, args []string) error {
-	client, err := api.ClientFromEnvironment()
+	client, err := api.FromEnv()
 	if err != nil {
 		return err
 	}
@@ -144,7 +144,7 @@ func RunHandler(cmd *cobra.Command, args []string) error {
 }

 func PushHandler(cmd *cobra.Command, args []string) error {
-	client, err := api.ClientFromEnvironment()
+	client, err := api.FromEnv()
 	if err != nil {
 		return err
 	}
@@ -188,7 +188,7 @@ func PushHandler(cmd *cobra.Command, args []string) error {
 }

 func ListHandler(cmd *cobra.Command, args []string) error {
-	client, err := api.ClientFromEnvironment()
+	client, err := api.FromEnv()
 	if err != nil {
 		return err
 	}
@@ -221,7 +221,7 @@ func ListHandler(cmd *cobra.Command, args []string) error {
 }

 func DeleteHandler(cmd *cobra.Command, args []string) error {
-	client, err := api.ClientFromEnvironment()
+	client, err := api.FromEnv()
 	if err != nil {
 		return err
 	}
@@ -237,7 +237,7 @@ func DeleteHandler(cmd *cobra.Command, args []string) error {
 }

 func ShowHandler(cmd *cobra.Command, args []string) error {
-	client, err := api.ClientFromEnvironment()
+	client, err := api.FromEnv()
 	if err != nil {
 		return err
 	}
@@ -315,7 +315,7 @@ func ShowHandler(cmd *cobra.Command, args []string) error {
 }

 func CopyHandler(cmd *cobra.Command, args []string) error {
-	client, err := api.ClientFromEnvironment()
+	client, err := api.FromEnv()
 	if err != nil {
 		return err
 	}
@@ -338,7 +338,7 @@ func PullHandler(cmd *cobra.Command, args []string) error {
 }

 func pull(model string, insecure bool) error {
-	client, err := api.ClientFromEnvironment()
+	client, err := api.FromEnv()
 	if err != nil {
 		return err
 	}
@@ -380,20 +380,7 @@ func pull(model string, insecure bool) error {
 func RunGenerate(cmd *cobra.Command, args []string) error {
 	if len(args) > 1 {
 		// join all args into a single prompt
-		wordWrap := false
-		if term.IsTerminal(int(os.Stdout.Fd())) {
-			wordWrap = true
-		}
-
-		nowrap, err := cmd.Flags().GetBool("nowordwrap")
-		if err != nil {
-			return err
-		}
-		if nowrap {
-			wordWrap = false
-		}
-
-		return generate(cmd, args[0], strings.Join(args[1:], " "), wordWrap)
+		return generate(cmd, args[0], strings.Join(args[1:], " "))
 	}

 	if readline.IsTerminal(int(os.Stdin.Fd())) {
@@ -405,8 +392,8 @@ func RunGenerate(cmd *cobra.Command, args []string) error {

 type generateContextKey string

-func generate(cmd *cobra.Command, model, prompt string, wordWrap bool) error {
-	client, err := api.ClientFromEnvironment()
+func generate(cmd *cobra.Command, model, prompt string) error {
+	client, err := api.FromEnv()
 	if err != nil {
 		return err
 	}
@@ -421,9 +408,24 @@ func generate(cmd *cobra.Command, model, prompt string, wordWrap bool) error {
 		generateContext = []int{}
 	}

+	var wrapTerm bool
+	termType := os.Getenv("TERM")
+	if termType == "xterm-256color" {
+		wrapTerm = true
+	}
+
 	termWidth, _, err := term.GetSize(int(0))
 	if err != nil {
-		wordWrap = false
+		wrapTerm = false
+	}
+
+	// override wrapping if the user turned it off
+	nowrap, err := cmd.Flags().GetBool("nowordwrap")
+	if err != nil {
+		return err
+	}
+	if nowrap {
+		wrapTerm = false
 	}

 	cancelCtx, cancel := context.WithCancel(context.Background())
@@ -450,7 +452,7 @@ func generate(cmd *cobra.Command, model, prompt string, wordWrap bool) error {

 		latest = response

-		if wordWrap {
+		if wrapTerm {
 			for _, ch := range response.Response {
 				if currentLineLength+1 > termWidth-5 {
 					// backtrack the length of the last word and clear to the end of the line
@@ -479,7 +481,18 @@ func generate(cmd *cobra.Command, model, prompt string, wordWrap bool) error {
 	}

 	if err := client.Generate(cancelCtx, &request, fn); err != nil {
-		if strings.Contains(err.Error(), "context canceled") && abort {
+		if strings.Contains(err.Error(), "failed to load model") {
+			// tell the user to check the server log, if it exists locally
+			home, nestedErr := os.UserHomeDir()
+			if nestedErr != nil {
+				// return the original error
+				return err
+			}
+			logPath := filepath.Join(home, ".ollama", "logs", "server.log")
+			if _, nestedErr := os.Stat(logPath); nestedErr == nil {
+				err = fmt.Errorf("%w\nFor more details, check the error logs at %s", err, logPath)
+			}
+		} else if strings.Contains(err.Error(), "context canceled") && abort {
 			spinner.Finish()
 			return nil
 		}
@@ -520,7 +533,7 @@ func generateInteractive(cmd *cobra.Command, model string) error {
 	}

 	// load the model
-	if err := generate(cmd, model, "", false); err != nil {
+	if err := generate(cmd, model, ""); err != nil {
 		return err
 	}

@@ -547,35 +560,8 @@ func generateInteractive(cmd *cobra.Command, model string) error {
 	)

 	usage := func() {
-		fmt.Fprintln(os.Stderr, "Available Commands:")
-		fmt.Fprintln(os.Stderr, "  /set         Set session variables")
-		fmt.Fprintln(os.Stderr, "  /show        Show model information")
-		fmt.Fprintln(os.Stderr, "  /bye         Exit")
-		fmt.Fprintln(os.Stderr, "  /?, /help    Help for a command")
-		fmt.Fprintln(os.Stderr, "")
-		fmt.Fprintln(os.Stderr, "Use \"\"\" to begin a multi-line message.")
-		fmt.Fprintln(os.Stderr, "")
-	}
-
-	usageSet := func() {
-		fmt.Fprintln(os.Stderr, "Available Commands:")
-		fmt.Fprintln(os.Stderr, "  /set history      Enable history")
-		fmt.Fprintln(os.Stderr, "  /set nohistory    Disable history")
-		fmt.Fprintln(os.Stderr, "  /set wordwrap     Enable wordwrap")
-		fmt.Fprintln(os.Stderr, "  /set nowordwrap   Disable wordwrap")
-		fmt.Fprintln(os.Stderr, "  /set verbose      Show LLM stats")
-		fmt.Fprintln(os.Stderr, "  /set quiet        Disable LLM stats")
-		fmt.Fprintln(os.Stderr, "")
-	}
-
-	usageShow := func() {
-		fmt.Fprintln(os.Stderr, "Available Commands:")
-		fmt.Fprintln(os.Stderr, "  /show license      Show model license")
-		fmt.Fprintln(os.Stderr, "  /show modelfile    Show Modelfile for this model")
-		fmt.Fprintln(os.Stderr, "  /show parameters   Show parameters for this model")
-		fmt.Fprintln(os.Stderr, "  /show system       Show system prompt")
-		fmt.Fprintln(os.Stderr, "  /show template     Show prompt template")
-		fmt.Fprintln(os.Stderr, "")
+		fmt.Fprintln(os.Stderr, "commands:")
+		fmt.Fprintln(os.Stderr, completer.Tree("  "))
 	}

 	var painter Painter
@@ -593,21 +579,6 @@ func generateInteractive(cmd *cobra.Command, model string) error {
 	}
 	defer scanner.Close()

-	var wordWrap bool
-	termType := os.Getenv("TERM")
-	if termType == "xterm-256color" {
-		wordWrap = true
-	}
-
-	// override wrapping if the user turned it off
-	nowrap, err := cmd.Flags().GetBool("nowordwrap")
-	if err != nil {
-		return err
-	}
-	if nowrap {
-		wordWrap = false
-	}
-
 	var multiLineBuffer string
 	var isMultiLine bool

@@ -661,10 +632,10 @@ func generateInteractive(cmd *cobra.Command, model string) error {
 				case "nohistory":
 					scanner.HistoryDisable()
 				case "wordwrap":
-					wordWrap = true
+					cmd.Flags().Set("nowordwrap", "false")
 					fmt.Println("Set 'wordwrap' mode.")
 				case "nowordwrap":
-					wordWrap = false
+					cmd.Flags().Set("nowordwrap", "true")
 					fmt.Println("Set 'nowordwrap' mode.")
 				case "verbose":
 					cmd.Flags().Set("verbose", "true")
@@ -689,7 +660,7 @@ func generateInteractive(cmd *cobra.Command, model string) error {
 					fmt.Printf("Unknown command '/set %s'. Type /? for help\n", args[1])
 				}
 			} else {
-				usageSet()
+				usage()
 			}
 		case strings.HasPrefix(line, "/show"):
 			args := strings.Fields(line)
@@ -702,49 +673,23 @@ func generateInteractive(cmd *cobra.Command, model string) error {

 				switch args[1] {
 				case "license":
-					if resp.License == "" {
-						fmt.Print("No license was specified for this model.\n\n")
-					} else {
-						fmt.Println(resp.License)
-					}
+					fmt.Println(resp.License)
 				case "modelfile":
 					fmt.Println(resp.Modelfile)
 				case "parameters":
-					if resp.Parameters == "" {
-						fmt.Print("No parameters were specified for this model.\n\n")
-					} else {
-						fmt.Println(resp.Parameters)
-					}
+					fmt.Println(resp.Parameters)
 				case "system":
-					if resp.System == "" {
-						fmt.Print("No system prompt was specified for this model.\n\n")
-					} else {
-						fmt.Println(resp.System)
-					}
+					fmt.Println(resp.System)
 				case "template":
-					if resp.Template == "" {
-						fmt.Print("No prompt template was specified for this model.\n\n")
-					} else {
-						fmt.Println(resp.Template)
-					}
+					fmt.Println(resp.Template)
 				default:
 					fmt.Printf("Unknown command '/show %s'. Type /? for help\n", args[1])
 				}
-			} else {
-				usageShow()
-			}
-		case strings.HasPrefix(line, "/help"), strings.HasPrefix(line, "/?"):
-			args := strings.Fields(line)
-			if len(args) > 1 {
-				switch args[1] {
-				case "set", "/set":
-					usageSet()
-				case "show", "/show":
-					usageShow()
-				}
 			} else {
 				usage()
 			}
+		case line == "/help", line == "/?":
+			usage()
 		case line == "/exit", line == "/bye":
 			return nil
 		case strings.HasPrefix(line, "/"):
@@ -753,7 +698,7 @@ func generateInteractive(cmd *cobra.Command, model string) error {
 		}

 		if len(line) > 0 && line[0] != '/' {
-			if err := generate(cmd, model, line, wordWrap); err != nil {
+			if err := generate(cmd, model, line); err != nil {
 				return err
 			}
 		}
@@ -765,7 +710,7 @@ func generateBatch(cmd *cobra.Command, model string) error {
 	for scanner.Scan() {
 		prompt := scanner.Text()
 		fmt.Printf(">>> %s\n", prompt)
-		if err := generate(cmd, model, prompt, false); err != nil {
+		if err := generate(cmd, model, prompt); err != nil {
 			return err
 		}
 	}
@@ -895,7 +840,7 @@ func startMacApp(client *api.Client) error {
 }

 func checkServerHeartbeat(_ *cobra.Command, _ []string) error {
-	client, err := api.ClientFromEnvironment()
+	client, err := api.FromEnv()
 	if err != nil {
 		return err
 	}
--- a/docs/api.md
+++ b/docs/api.md
@@ -12,6 +12,7 @@
 - [Push a Model](#push-a-model)
 - [Generate Embeddings](#generate-embeddings)

+
 ## Conventions

 ### Model names
@@ -39,13 +40,12 @@ Generate a response for a given prompt with a provided model. This is a streamin
 - `model`: (required) the [model name](#model-names)
 - `prompt`: the prompt to generate a response for

-Advanced parameters (optional):
+Advanced parameters:

 - `options`: additional model parameters listed in the documentation for the [Modelfile](./modelfile.md#valid-parameters-and-values) such as `temperature`
 - `system`: system prompt to (overrides what is defined in the `Modelfile`)
 - `template`: the full prompt or prompt template (overrides what is defined in the `Modelfile`)
 - `context`: the context parameter returned from a previous request to `/generate`, this can be used to keep a short conversational memory
- `stream`: if `false` the response will be be returned as a single response object, rather than a stream of objects

 ### Request

@@ -80,7 +80,6 @@ The final response in the stream also includes additional data about the generat
 - `eval_count`: number of tokens the response
 - `eval_duration`: time in nanoseconds spent generating the response
 - `context`: an encoding of the conversation used in this response, this can be sent in the next request to keep a conversational memory
- `response`: empty if the response was streamed, if not streamed, this will contain the full response

 To calculate how fast the response is generated in tokens per second (token/s), divide `eval_count` / `eval_duration`.

@@ -88,7 +87,6 @@ To calculate how fast the response is generated in tokens per second (token/s),
 {
  "model": "llama2:7b",
  "created_at": "2023-08-04T19:22:45.499127Z",
-  "response": "",
  "context": [1, 2, 3],
  "done": true,
  "total_duration": 5589157167,
@@ -114,7 +112,6 @@ Create a model from a [`Modelfile`](./modelfile.md)

 - `name`: name of the model to create
 - `path`: path to the Modelfile
- `stream`: (optional) if `false` the response will be be returned as a single response object, rather than a stream of objects

 ### Request

@@ -182,7 +179,7 @@ Show details about a model including modelfile, template, parameters, license, a

 ### Request

-```shell
+```shell  
 curl http://localhost:11434/api/show -d '{
  "name": "llama2:7b"
 }'
@@ -192,10 +189,10 @@ curl http://localhost:11434/api/show -d '{

 ```json
 {
-  "license": "<contents of license block>",
-  "modelfile": "# Modelfile generated by \"ollama show\"\n# To build a new Modelfile based on this one, replace the FROM line with:\n# FROM llama2:latest\n\nFROM /Users/username/.ollama/models/blobs/sha256:8daa9615cce30c259a9555b1cc250d461d1bc69980a274b44d7eda0be78076d8\nTEMPLATE \"\"\"[INST] {{ if and .First .System }}<<SYS>>{{ .System }}<</SYS>>\n\n{{ end }}{{ .Prompt }} [/INST] \"\"\"\nSYSTEM \"\"\"\"\"\"\nPARAMETER stop [INST]\nPARAMETER stop [/INST]\nPARAMETER stop <<SYS>>\nPARAMETER stop <</SYS>>\n",
-  "parameters": "stop                           [INST]\nstop                           [/INST]\nstop                           <<SYS>>\nstop                           <</SYS>>",
-  "template": "[INST] {{ if and .First .System }}<<SYS>>{{ .System }}<</SYS>>\n\n{{ end }}{{ .Prompt }} [/INST] "
+    "license": "<contents of license block>",
+    "modelfile": "# Modelfile generated by \"ollama show\"\n# To build a new Modelfile based on this one, replace the FROM line with:\n# FROM llama2:latest\n\nFROM /Users/username/.ollama/models/blobs/sha256:8daa9615cce30c259a9555b1cc250d461d1bc69980a274b44d7eda0be78076d8\nTEMPLATE \"\"\"[INST] {{ if and .First .System }}<<SYS>>{{ .System }}<</SYS>>\n\n{{ end }}{{ .Prompt }} [/INST] \"\"\"\nSYSTEM \"\"\"\"\"\"\nPARAMETER stop [INST]\nPARAMETER stop [/INST]\nPARAMETER stop <<SYS>>\nPARAMETER stop <</SYS>>\n",
+    "parameters": "stop                           [INST]\nstop                           [/INST]\nstop                           <<SYS>>\nstop                           <</SYS>>",
+    "template": "[INST] {{ if and .First .System }}<<SYS>>{{ .System }}<</SYS>>\n\n{{ end }}{{ .Prompt }} [/INST] "
 }
 ```

@@ -248,7 +245,6 @@ Download a model from the ollama library. Cancelled pulls are resumed from where

 - `name`: name of the model to pull
 - `insecure`: (optional) allow insecure connections to the library. Only use this if you are pulling from your own library during development.
- `stream`: (optional) if `false` the response will be be returned as a single response object, rather than a stream of objects

 ### Request

@@ -279,8 +275,7 @@ Upload a model to a model library. Requires registering for ollama.ai and adding
 ### Parameters

 - `name`: name of the model to push in the form of `<namespace>/<model>:<tag>`
- `insecure`: (optional) allow insecure connections to the library. Only use this if you are pushing to your library during development.
- `stream`: (optional) if `false` the response will be be returned as a single response object, rather than a stream of objects
+- `insecure`: (optional) allow insecure connections to the library. Only use this if you are pushing to your library during development.  

 ### Request

@@ -295,16 +290,15 @@ curl -X POST http://localhost:11434/api/push -d '{
 Streaming response that starts with:

 ```json
-{ "status": "retrieving manifest" }
+{"status":"retrieving manifest"}
 ```

 and then:

 ```json
 {
-  "status": "starting upload",
-  "digest": "sha256:bc07c81de745696fdf5afca05e065818a8149fb0c77266fb584d9b2cba3711ab",
-  "total": 1928429856
+"status":"starting upload","digest":"sha256:bc07c81de745696fdf5afca05e065818a8149fb0c77266fb584d9b2cba3711ab",
+"total":1928429856
 }
 ```

@@ -312,10 +306,9 @@ Then there is a series of uploading responses:

 ```json
 {
-  "status": "starting upload",
-  "digest": "sha256:bc07c81de745696fdf5afca05e065818a8149fb0c77266fb584d9b2cba3711ab",
-  "total": 1928429856
-}
+"status":"starting upload",
+"digest":"sha256:bc07c81de745696fdf5afca05e065818a8149fb0c77266fb584d9b2cba3711ab",
+"total":1928429856}
 ```

 Finally, when the upload is complete:
@@ -359,5 +352,4 @@ curl -X POST http://localhost:11434/api/embeddings -d '{
    0.5670403838157654, 0.009260174818336964, 0.23178744316101074, -0.2916173040866852, -0.8924556970596313,
    0.8785552978515625, -0.34576427936553955, 0.5742510557174683, -0.04222835972905159, -0.137906014919281
  ]
-}
-```
+}```
--- a/docs/development.md
+++ b/docs/development.md
@@ -10,25 +10,25 @@ Install required tools:
 - go version 1.20 or higher
 - gcc version 11.4.0 or higher

-```bash
+```
 brew install go cmake gcc
 ```

 Get the required libraries:

-```bash
+```
 go generate ./...
 ```

 Then build ollama:

-```bash
+```
 go build .
 ```

 Now you can run `ollama`:

-```bash
+```
 ./ollama
 ```

--- a/docs/faq.md
+++ b/docs/faq.md
@@ -2,13 +2,13 @@

 ## How can I expose the Ollama server?

-```bash
+```
 OLLAMA_HOST=0.0.0.0:11435 ollama serve
 ```

 By default, Ollama allows cross origin requests from `127.0.0.1` and `0.0.0.0`. To support more origins, you can use the `OLLAMA_ORIGINS` environment variable:

-```bash
+```
 OLLAMA_ORIGINS=http://192.168.1.1:*,https://example.com ollama serve
 ```

@@ -16,3 +16,4 @@ OLLAMA_ORIGINS=http://192.168.1.1:*,https://example.com ollama serve

 * macOS: Raw model data is stored under `~/.ollama/models`.
 * Linux: Raw model data is stored under `/usr/share/ollama/.ollama/models`
+
--- a/docs/linux.md
+++ b/docs/linux.md
@@ -2,7 +2,7 @@

 > Note: A one line installer for Ollama is available by running:
 >
-> ```bash
+> ```
 > curl https://ollama.ai/install.sh | sh
 > ```

@@ -10,7 +10,7 @@

 Ollama is distributed as a self-contained binary. Download it to a directory in your PATH:

-```bash
+```
 sudo curl -L https://ollama.ai/download/ollama-linux-amd64 -o /usr/bin/ollama
 sudo chmod +x /usr/bin/ollama
 ```
@@ -19,13 +19,13 @@ sudo chmod +x /usr/bin/ollama

 Start Ollama by running `ollama serve`:

-```bash
+```
 ollama serve
 ```

 Once Ollama is running, run a model in another terminal session:

-```bash
+```
 ollama run llama2
 ```

@@ -35,7 +35,7 @@ ollama run llama2

 Verify that the drivers are installed by running the following command, which should print details about your GPU:

-```bash
+```
 nvidia-smi
 ```

@@ -43,7 +43,7 @@ nvidia-smi

 Create a user for Ollama:

-```bash
+```
 sudo useradd -r -s /bin/false -m -d /usr/share/ollama ollama
 ```

@@ -68,7 +68,7 @@ WantedBy=default.target

 Then start the service:

-```bash
+```
 sudo systemctl daemon-reload
 sudo systemctl enable ollama
 ```
@@ -77,7 +77,7 @@ sudo systemctl enable ollama

 To view logs of Ollama running as a startup service, run:

-```bash
+```
 journalctl -u ollama
 ```

--- a/docs/modelfile.md
+++ b/docs/modelfile.md
@@ -1,6 +1,6 @@
 # Ollama Model File

-> Note: this `Modelfile` syntax is in development
+> Note: this model file syntax is in development

 A model file is the blueprint to create and share models with Ollama.

@@ -24,7 +24,7 @@ A model file is the blueprint to create and share models with Ollama.

 ## Format

-The format of the `Modelfile`:
+The format of the Modelfile:

 ```modelfile
 # comment
@@ -42,9 +42,9 @@ INSTRUCTION arguments

 ## Examples

-An example of a `Modelfile` creating a mario blueprint:
+An example of a model file creating a mario blueprint:

-```modelfile
+```
 FROM llama2
 # sets the temperature to 1 [higher is more creative, lower is more coherent]
 PARAMETER temperature 1
@@ -57,9 +57,9 @@ SYSTEM You are Mario from super mario bros, acting as an assistant.

 To use this:

-1. Save it as a file (e.g. `Modelfile`)
-2. `ollama create choose-a-model-name -f <location of the file e.g. ./Modelfile>'`
-3. `ollama run choose-a-model-name`
+1. Save it as a file (eg. `Modelfile`)
+2. `ollama create NAME -f <location of the file eg. ./Modelfile>'`
+3. `ollama run NAME`
 4. Start using the model!

 More examples are available in the [examples directory](../examples).
@@ -68,34 +68,34 @@ More examples are available in the [examples directory](../examples).

 ### FROM (Required)

-The `FROM` instruction defines the base model to use when creating a model.
+The FROM instruction defines the base model to use when creating a model.

-```modelfile
+```
 FROM <model name>:<tag>
 ```

 #### Build from llama2

-```modelfile
+```
 FROM llama2
 ```

 A list of available base models:
 <https://github.com/jmorganca/ollama#model-library>

-#### Build from a `bin` file
+#### Build from a bin file

-```modelfile
+```
 FROM ./ollama-model.bin
 ```

-This bin file location should be specified as an absolute path or relative to the `Modelfile` location.
+This bin file location should be specified as an absolute path or relative to the Modelfile location.

 ### EMBED

-The `EMBED` instruction is used to add embeddings of files to a model. This is useful for adding custom data that the model can reference when generating an answer. Note that currently only text files are supported, formatted with each line as one embedding.
+The EMBED instruction is used to add embeddings of files to a model. This is useful for adding custom data that the model can reference when generating an answer. Note that currently only text files are supported, formatted with each line as one embedding.

-```modelfile
+```
 FROM <model name>:<tag>
 EMBED <file path>.txt
 EMBED <different file path>.txt
@@ -106,7 +106,7 @@ EMBED <path to directory>/*.txt

 The `PARAMETER` instruction defines a parameter that can be set when the model is run.

-```modelfile
+```
 PARAMETER <parameter> <parametervalue>
 ```

@@ -142,7 +142,7 @@ PARAMETER <parameter> <parametervalue>
 | `{{ .Prompt }}` | The incoming prompt, this is not specified in the model file and will be set based on input.                 |
 | `{{ .First }}`  | A boolean value used to render specific template information for the first generation of a session.          |

-```modelfile
+```
 TEMPLATE """
 {{- if .First }}
 ### System:
@@ -162,7 +162,7 @@ SYSTEM """<system message>"""

 The `SYSTEM` instruction specifies the system prompt to be used in the template, if applicable.

-```modelfile
+```
 SYSTEM """<system message>"""
 ```

@@ -170,7 +170,7 @@ SYSTEM """<system message>"""

 The `ADAPTER` instruction specifies the LoRA adapter to apply to the base model. The value of this instruction should be an absolute path or a path relative to the Modelfile and the file must be in a GGML file format. The adapter should be tuned from the base model otherwise the behaviour is undefined.

-```modelfile
+```
 ADAPTER ./ollama-lora.bin
 ```

@@ -178,7 +178,7 @@ ADAPTER ./ollama-lora.bin

 The `LICENSE` instruction allows you to specify the legal license under which the model used with this Modelfile is shared or distributed.

-```modelfile
+```
 LICENSE """
 <license text>
 """
@@ -186,5 +186,5 @@ LICENSE """

 ## Notes

- the **`Modelfile` is not case sensitive**. In the examples, we use uppercase for instructions to make it easier to distinguish it from arguments.
+- the **modelfile is not case sensitive**. In the examples, we use uppercase for instructions to make it easier to distinguish it from arguments.
 - Instructions can be in any order. In the examples, we start with FROM instruction to keep it easily readable.
--- a/examples/.gitignore
+++ b/examples/.gitignore
@@ -1,171 +0,0 @@
-node_modules
-# OSX
-.DS_STORE
-
-# Models
-models/
-
-# Local Chroma db
-.chroma/
-db/
-
-# Byte-compiled / optimized / DLL files
-__pycache__/
-*.py[cod]
-*$py.class
-
-# C extensions
-*.so
-
-# Distribution / packaging
-.Python
-build/
-develop-eggs/
-dist/
-downloads/
-eggs/
-.eggs/
-lib/
-lib64/
-parts/
-sdist/
-var/
-wheels/
-share/python-wheels/
-*.egg-info/
-.installed.cfg
-*.egg
-MANIFEST
-
-# PyInstaller
-#  Usually these files are written by a python script from a template
-#  before PyInstaller builds the exe, so as to inject date/other infos into it.
-*.manifest
-*.spec
-
-# Installer logs
-pip-log.txt
-pip-delete-this-directory.txt
-
-# Unit test / coverage reports
-htmlcov/
-.tox/
-.nox/
-.coverage
-.coverage.*
-.cache
-nosetests.xml
-coverage.xml
-*.cover
-*.py,cover
-.hypothesis/
-.pytest_cache/
-cover/
-
-# Translations
-*.mo
-*.pot
-
-# Django stuff:
-*.log
-local_settings.py
-db.sqlite3
-db.sqlite3-journal
-
-# Flask stuff:
-instance/
-.webassets-cache
-
-# Scrapy stuff:
-.scrapy
-
-# Sphinx documentation
-docs/_build/
-
-# PyBuilder
-.pybuilder/
-target/
-
-# Jupyter Notebook
-.ipynb_checkpoints
-
-# IPython
-profile_default/
-ipython_config.py
-
-# pyenv
-#   For a library or package, you might want to ignore these files since the code is
-#   intended to run in multiple environments; otherwise, check them in:
-# .python-version
-
-# pipenv
-#   According to pypa/pipenv#598, it is recommended to include Pipfile.lock in version control.
-#   However, in case of collaboration, if having platform-specific dependencies or dependencies
-#   having no cross-platform support, pipenv may install dependencies that don't work, or not
-#   install all needed dependencies.
-#Pipfile.lock
-
-# poetry
-#   Similar to Pipfile.lock, it is generally recommended to include poetry.lock in version control.
-#   This is especially recommended for binary packages to ensure reproducibility, and is more
-#   commonly ignored for libraries.
-#   https://python-poetry.org/docs/basic-usage/#commit-your-poetrylock-file-to-version-control
-#poetry.lock
-
-# pdm
-#   Similar to Pipfile.lock, it is generally recommended to include pdm.lock in version control.
-#pdm.lock
-#   pdm stores project-wide configurations in .pdm.toml, but it is recommended to not include it
-#   in version control.
-#   https://pdm.fming.dev/#use-with-ide
-.pdm.toml
-
-# PEP 582; used by e.g. github.com/David-OConnor/pyflow and github.com/pdm-project/pdm
-__pypackages__/
-
-# Celery stuff
-celerybeat-schedule
-celerybeat.pid
-
-# SageMath parsed files
-*.sage.py
-
-# Environments
-.env
-.venv
-env/
-venv/
-ENV/
-env.bak/
-venv.bak/
-
-# Spyder project settings
-.spyderproject
-.spyproject
-
-# Rope project settings
-.ropeproject
-
-# mkdocs documentation
-/site
-
-# mypy
-.mypy_cache/
-.dmypy.json
-dmypy.json
-
-# Pyre type checker
-.pyre/
-
-# pytype static type analyzer
-.pytype/
-
-# Cython debug symbols
-cython_debug/
-
-# PyCharm
-#  JetBrains specific template is maintained in a separate JetBrains.gitignore that can
-#  be found at https://github.com/github/gitignore/blob/main/Global/JetBrains.gitignore
-#  and can be added to the global gitignore or merged into this file.  For a more nuclear
-#  option (not recommended) you can uncomment the following to ignore the entire idea folder.
-#.idea/
--- a/examples/modelfile-10tweets/Modelfile
+++ b/examples/modelfile-10tweets/Modelfile
--- a/examples/README.md
+++ b/examples/README.md
@@ -1,3 +1,15 @@
 # Examples

-This directory contains different examples of using Ollama.
+This directory contains different examples of using Ollama
+
+To create a model:
+
+```
+ollama create example -f <example file>
+```
+
+To run a model:
+
+```
+ollama run example
+```
--- a/examples/modelfile-devopsengineer/Modelfile
+++ b/examples/modelfile-devopsengineer/Modelfile
@@ -1,7 +1,7 @@
 # Modelfile for creating a devops engineer assistant
 # Run `ollama create devops-engineer -f ./Modelfile` and then `ollama run devops-engineer` and enter a topic

-FROM mistral
+FROM llama2:13b
 PARAMETER temperature 1
 SYSTEM """
 You are a senior devops engineer, acting as an assistant. You offer help with cloud technologies like: Terraform, AWS, kubernetes, python. You answer with code examples when possible
--- a/examples/python-dockerit/Modelfile
+++ b/examples/python-dockerit/Modelfile
@@ -1,4 +1,4 @@
-FROM mistral
+FROM llama2
 SYSTEM """
 You are an experienced Devops engineer focused on docker. When given specifications for a particular need or application you know the best way to host that within a docker container. For instance if someone tells you they want an nginx server to host files located at /web you will answer as follows

--- a/examples/python-dockerit/README.md
+++ b/examples/python-dockerit/README.md
--- a/examples/python-dockerit/dockerit.py
+++ b/examples/python-dockerit/dockerit.py
--- a/examples/python-dockerit/requirements.txt
+++ b/examples/python-dockerit/requirements.txt
--- a/examples/golang-simplegenerate/README.md
+++ b/examples/golang-simplegenerate/README.md
--- a/examples/golang-simplegenerate/main.go
+++ b/examples/golang-simplegenerate/main.go
@@ -1,27 +0,0 @@
-package main
-
-import (
-	"bytes"
-	"fmt"
-	"net/http"
-	"os"
-	"io"
-	"log"
-)
-
-func main() {
-	body := []byte(`{"model":"mistral"}`)
-	resp, err := http.Post("http://localhost:11434/api/generate", "application/json", bytes.NewBuffer(body))
-
-	if err != nil {
-		fmt.Print(err.Error())
-		os.Exit(1)
-	} 
-
-	responseData, err := io.ReadAll(resp.Body)
-	if err != nil {
-		log.Fatal(err)
-	}
-	fmt.Println(string(responseData))
-
-}
--- a/examples/langchain-python-rag-document/README.md
+++ b/examples/langchain-python-rag-document/README.md
--- a/examples/langchain-python-rag-document/main.py
+++ b/examples/langchain-python-rag-document/main.py
--- a/examples/langchain-python-rag-document/requirements.txt
+++ b/examples/langchain-python-rag-document/requirements.txt
--- a/examples/langchain-typescript-simple/README.md
+++ b/examples/langchain-typescript-simple/README.md
@@ -1,21 +0,0 @@
-# LangChain
-
-This example is a basic "hello world" of using LangChain with Ollama using Node.js and Typescript.
-
-## Setup
-
-```shell
-npm install
-```
-
-## Run
-
-```shell
-ts-node main.ts
-```
-
-Running this example will print the response for "hello":
-
-```plaintext
-Hello! It's nice to meet you. hopefully you are having a great day! Is there something I can help you with or would you like to chat?
-```
--- a/examples/langchain-typescript-simple/main.ts
+++ b/examples/langchain-typescript-simple/main.ts
@@ -1,15 +0,0 @@
-import { Ollama} from 'langchain/llms/ollama';
-
-async function main() {
-  const ollama = new Ollama({
-    model: 'mistral'    
-    // other parameters can be found at https://js.langchain.com/docs/api/llms_ollama/classes/Ollama
-  })
-  const stream = await ollama.stream("Hello");
-
-  for await (const chunk of stream) {
-    process.stdout.write(chunk);
-  }
-}
-
-main();
--- a/examples/langchain-typescript-simple/package-lock.json
+++ b/examples/langchain-typescript-simple/package-lock.json
@@ -1,997 +0,0 @@
-{
-  "name": "with-langchain-typescript-simplegenerate",
-  "lockfileVersion": 3,
-  "requires": true,
-  "packages": {
-    "": {
-      "dependencies": {
-        "langchain": "^0.0.165"
-      },
-      "devDependencies": {
-        "typescript": "^5.2.2"
-      }
-    },
-    "node_modules/@anthropic-ai/sdk": {
-      "version": "0.6.2",
-      "resolved": "https://registry.npmjs.org/@anthropic-ai/sdk/-/sdk-0.6.2.tgz",
-      "integrity": "sha512-fB9PUj9RFT+XjkL+E9Ol864ZIJi+1P8WnbHspN3N3/GK2uSzjd0cbVIKTGgf4v3N8MwaQu+UWnU7C4BG/fap/g==",
-      "dependencies": {
-        "@types/node": "^18.11.18",
-        "@types/node-fetch": "^2.6.4",
-        "abort-controller": "^3.0.0",
-        "agentkeepalive": "^4.2.1",
-        "digest-fetch": "^1.3.0",
-        "form-data-encoder": "1.7.2",
-        "formdata-node": "^4.3.2",
-        "node-fetch": "^2.6.7"
-      }
-    },
-    "node_modules/@types/node": {
-      "version": "18.18.4",
-      "resolved": "https://registry.npmjs.org/@types/node/-/node-18.18.4.tgz",
-      "integrity": "sha512-t3rNFBgJRugIhackit2mVcLfF6IRc0JE4oeizPQL8Zrm8n2WY/0wOdpOPhdtG0V9Q2TlW/axbF1MJ6z+Yj/kKQ=="
-    },
-    "node_modules/@types/node-fetch": {
-      "version": "2.6.6",
-      "resolved": "https://registry.npmjs.org/@types/node-fetch/-/node-fetch-2.6.6.tgz",
-      "integrity": "sha512-95X8guJYhfqiuVVhRFxVQcf4hW/2bCuoPwDasMf/531STFoNoWTT7YDnWdXHEZKqAGUigmpG31r2FE70LwnzJw==",
-      "dependencies": {
-        "@types/node": "*",
-        "form-data": "^4.0.0"
-      }
-    },
-    "node_modules/@types/retry": {
-      "version": "0.12.0",
-      "resolved": "https://registry.npmjs.org/@types/retry/-/retry-0.12.0.tgz",
-      "integrity": "sha512-wWKOClTTiizcZhXnPY4wikVAwmdYHp8q6DmC+EJUzAMsycb7HB32Kh9RN4+0gExjmPmZSAQjgURXIGATPegAvA=="
-    },
-    "node_modules/@types/uuid": {
-      "version": "9.0.5",
-      "resolved": "https://registry.npmjs.org/@types/uuid/-/uuid-9.0.5.tgz",
-      "integrity": "sha512-xfHdwa1FMJ082prjSJpoEI57GZITiQz10r3vEJCHa2khEFQjKy91aWKz6+zybzssCvXUwE1LQWgWVwZ4nYUvHQ=="
-    },
-    "node_modules/abort-controller": {
-      "version": "3.0.0",
-      "resolved": "https://registry.npmjs.org/abort-controller/-/abort-controller-3.0.0.tgz",
-      "integrity": "sha512-h8lQ8tacZYnR3vNQTgibj+tODHI5/+l06Au2Pcriv/Gmet0eaj4TwWH41sO9wnHDiQsEj19q0drzdWdeAHtweg==",
-      "dependencies": {
-        "event-target-shim": "^5.0.0"
-      },
-      "engines": {
-        "node": ">=6.5"
-      }
-    },
-    "node_modules/agentkeepalive": {
-      "version": "4.5.0",
-      "resolved": "https://registry.npmjs.org/agentkeepalive/-/agentkeepalive-4.5.0.tgz",
-      "integrity": "sha512-5GG/5IbQQpC9FpkRGsSvZI5QYeSCzlJHdpBQntCsuTOxhKD8lqKhrleg2Yi7yvMIf82Ycmmqln9U8V9qwEiJew==",
-      "dependencies": {
-        "humanize-ms": "^1.2.1"
-      },
-      "engines": {
-        "node": ">= 8.0.0"
-      }
-    },
-    "node_modules/ansi-styles": {
-      "version": "5.2.0",
-      "resolved": "https://registry.npmjs.org/ansi-styles/-/ansi-styles-5.2.0.tgz",
-      "integrity": "sha512-Cxwpt2SfTzTtXcfOlzGEee8O+c+MmUgGrNiBcXnuWxuFJHe6a5Hz7qwhwe5OgaSYI0IJvkLqWX1ASG+cJOkEiA==",
-      "engines": {
-        "node": ">=10"
-      },
-      "funding": {
-        "url": "https://github.com/chalk/ansi-styles?sponsor=1"
-      }
-    },
-    "node_modules/argparse": {
-      "version": "2.0.1",
-      "resolved": "https://registry.npmjs.org/argparse/-/argparse-2.0.1.tgz",
-      "integrity": "sha512-8+9WqebbFzpX9OR+Wa6O29asIogeRMzcGtAINdpMHHyAg10f05aSFVBbcEqGf/PXw1EjAZ+q2/bEBg3DvurK3Q=="
-    },
-    "node_modules/asynckit": {
-      "version": "0.4.0",
-      "resolved": "https://registry.npmjs.org/asynckit/-/asynckit-0.4.0.tgz",
-      "integrity": "sha512-Oei9OH4tRh0YqU3GxhX79dM/mwVgvbZJaSNaRk+bshkj0S5cfHcgYakreBjrHwatXKbz+IoIdYLxrKim2MjW0Q=="
-    },
-    "node_modules/base-64": {
-      "version": "0.1.0",
-      "resolved": "https://registry.npmjs.org/base-64/-/base-64-0.1.0.tgz",
-      "integrity": "sha512-Y5gU45svrR5tI2Vt/X9GPd3L0HNIKzGu202EjxrXMpuc2V2CiKgemAbUUsqYmZJvPtCXoUKjNZwBJzsNScUbXA=="
-    },
-    "node_modules/base64-js": {
-      "version": "1.5.1",
-      "resolved": "https://registry.npmjs.org/base64-js/-/base64-js-1.5.1.tgz",
-      "integrity": "sha512-AKpaYlHn8t4SVbOHCy+b5+KKgvR4vrsD8vbvrbiQJps7fKDTkjkDry6ji0rUJjC0kzbNePLwzxq8iypo41qeWA==",
-      "funding": [
-        {
-          "type": "github",
-          "url": "https://github.com/sponsors/feross"
-        },
-        {
-          "type": "patreon",
-          "url": "https://www.patreon.com/feross"
-        },
-        {
-          "type": "consulting",
-          "url": "https://feross.org/support"
-        }
-      ]
-    },
-    "node_modules/binary-extensions": {
-      "version": "2.2.0",
-      "resolved": "https://registry.npmjs.org/binary-extensions/-/binary-extensions-2.2.0.tgz",
-      "integrity": "sha512-jDctJ/IVQbZoJykoeHbhXpOlNBqGNcwXJKJog42E5HDPUwQTSdjCHdihjj0DlnheQ7blbT6dHOafNAiS8ooQKA==",
-      "engines": {
-        "node": ">=8"
-      }
-    },
-    "node_modules/binary-search": {
-      "version": "1.3.6",
-      "resolved": "https://registry.npmjs.org/binary-search/-/binary-search-1.3.6.tgz",
-      "integrity": "sha512-nbE1WxOTTrUWIfsfZ4aHGYu5DOuNkbxGokjV6Z2kxfJK3uaAb8zNK1muzOeipoLHZjInT4Br88BHpzevc681xA=="
-    },
-    "node_modules/camelcase": {
-      "version": "6.3.0",
-      "resolved": "https://registry.npmjs.org/camelcase/-/camelcase-6.3.0.tgz",
-      "integrity": "sha512-Gmy6FhYlCY7uOElZUSbxo2UCDH8owEk996gkbrpsgGtrJLM3J7jGxl9Ic7Qwwj4ivOE5AWZWRMecDdF7hqGjFA==",
-      "engines": {
-        "node": ">=10"
-      },
-      "funding": {
-        "url": "https://github.com/sponsors/sindresorhus"
-      }
-    },
-    "node_modules/charenc": {
-      "version": "0.0.2",
-      "resolved": "https://registry.npmjs.org/charenc/-/charenc-0.0.2.tgz",
-      "integrity": "sha512-yrLQ/yVUFXkzg7EDQsPieE/53+0RlaWTs+wBrvW36cyilJ2SaDWfl4Yj7MtLTXleV9uEKefbAGUPv2/iWSooRA==",
-      "engines": {
-        "node": "*"
-      }
-    },
-    "node_modules/combined-stream": {
-      "version": "1.0.8",
-      "resolved": "https://registry.npmjs.org/combined-stream/-/combined-stream-1.0.8.tgz",
-      "integrity": "sha512-FQN4MRfuJeHf7cBbBMJFXhKSDq+2kAArBlmRBvcvFE5BB1HZKXtSFASDhdlz9zOYwxh8lDdnvmMOe/+5cdoEdg==",
-      "dependencies": {
-        "delayed-stream": "~1.0.0"
-      },
-      "engines": {
-        "node": ">= 0.8"
-      }
-    },
-    "node_modules/commander": {
-      "version": "10.0.1",
-      "resolved": "https://registry.npmjs.org/commander/-/commander-10.0.1.tgz",
-      "integrity": "sha512-y4Mg2tXshplEbSGzx7amzPwKKOCGuoSRP/CjEdwwk0FOGlUbq6lKuoyDZTNZkmxHdJtp54hdfY/JUrdL7Xfdug==",
-      "engines": {
-        "node": ">=14"
-      }
-    },
-    "node_modules/crypt": {
-      "version": "0.0.2",
-      "resolved": "https://registry.npmjs.org/crypt/-/crypt-0.0.2.tgz",
-      "integrity": "sha512-mCxBlsHFYh9C+HVpiEacem8FEBnMXgU9gy4zmNC+SXAZNB/1idgp/aulFJ4FgCi7GPEVbfyng092GqL2k2rmow==",
-      "engines": {
-        "node": "*"
-      }
-    },
-    "node_modules/decamelize": {
-      "version": "1.2.0",
-      "resolved": "https://registry.npmjs.org/decamelize/-/decamelize-1.2.0.tgz",
-      "integrity": "sha512-z2S+W9X73hAUUki+N+9Za2lBlun89zigOyGrsax+KUQ6wKW4ZoWpEYBkGhQjwAjjDCkWxhY0VKEhk8wzY7F5cA==",
-      "engines": {
-        "node": ">=0.10.0"
-      }
-    },
-    "node_modules/delayed-stream": {
-      "version": "1.0.0",
-      "resolved": "https://registry.npmjs.org/delayed-stream/-/delayed-stream-1.0.0.tgz",
-      "integrity": "sha512-ZySD7Nf91aLB0RxL4KGrKHBXl7Eds1DAmEdcoVawXnLD7SDhpNgtuII2aAkg7a7QS41jxPSZ17p4VdGnMHk3MQ==",
-      "engines": {
-        "node": ">=0.4.0"
-      }
-    },
-    "node_modules/digest-fetch": {
-      "version": "1.3.0",
-      "resolved": "https://registry.npmjs.org/digest-fetch/-/digest-fetch-1.3.0.tgz",
-      "integrity": "sha512-CGJuv6iKNM7QyZlM2T3sPAdZWd/p9zQiRNS9G+9COUCwzWFTs0Xp8NF5iePx7wtvhDykReiRRrSeNb4oMmB8lA==",
-      "dependencies": {
-        "base-64": "^0.1.0",
-        "md5": "^2.3.0"
-      }
-    },
-    "node_modules/event-target-shim": {
-      "version": "5.0.1",
-      "resolved": "https://registry.npmjs.org/event-target-shim/-/event-target-shim-5.0.1.tgz",
-      "integrity": "sha512-i/2XbnSz/uxRCU6+NdVJgKWDTM427+MqYbkQzD321DuCQJUqOuJKIA0IM2+W2xtYHdKOmZ4dR6fExsd4SXL+WQ==",
-      "engines": {
-        "node": ">=6"
-      }
-    },
-    "node_modules/eventemitter3": {
-      "version": "4.0.7",
-      "resolved": "https://registry.npmjs.org/eventemitter3/-/eventemitter3-4.0.7.tgz",
-      "integrity": "sha512-8guHBZCwKnFhYdHr2ysuRWErTwhoN2X8XELRlrRwpmfeY2jjuUN4taQMsULKUVo1K4DvZl+0pgfyoysHxvmvEw=="
-    },
-    "node_modules/expr-eval": {
-      "version": "2.0.2",
-      "resolved": "https://registry.npmjs.org/expr-eval/-/expr-eval-2.0.2.tgz",
-      "integrity": "sha512-4EMSHGOPSwAfBiibw3ndnP0AvjDWLsMvGOvWEZ2F96IGk0bIVdjQisOHxReSkE13mHcfbuCiXw+G4y0zv6N8Eg=="
-    },
-    "node_modules/flat": {
-      "version": "5.0.2",
-      "resolved": "https://registry.npmjs.org/flat/-/flat-5.0.2.tgz",
-      "integrity": "sha512-b6suED+5/3rTpUBdG1gupIl8MPFCAMA0QXwmljLhvCUKcUvdE4gWky9zpuGCcXHOsz4J9wPGNWq6OKpmIzz3hQ==",
-      "bin": {
-        "flat": "cli.js"
-      }
-    },
-    "node_modules/form-data": {
-      "version": "4.0.0",
-      "resolved": "https://registry.npmjs.org/form-data/-/form-data-4.0.0.tgz",
-      "integrity": "sha512-ETEklSGi5t0QMZuiXoA/Q6vcnxcLQP5vdugSpuAyi6SVGi2clPPp+xgEhuMaHC+zGgn31Kd235W35f7Hykkaww==",
-      "dependencies": {
-        "asynckit": "^0.4.0",
-        "combined-stream": "^1.0.8",
-        "mime-types": "^2.1.12"
-      },
-      "engines": {
-        "node": ">= 6"
-      }
-    },
-    "node_modules/form-data-encoder": {
-      "version": "1.7.2",
-      "resolved": "https://registry.npmjs.org/form-data-encoder/-/form-data-encoder-1.7.2.tgz",
-      "integrity": "sha512-qfqtYan3rxrnCk1VYaA4H+Ms9xdpPqvLZa6xmMgFvhO32x7/3J/ExcTd6qpxM0vH2GdMI+poehyBZvqfMTto8A=="
-    },
-    "node_modules/formdata-node": {
-      "version": "4.4.1",
-      "resolved": "https://registry.npmjs.org/formdata-node/-/formdata-node-4.4.1.tgz",
-      "integrity": "sha512-0iirZp3uVDjVGt9p49aTaqjk84TrglENEDuqfdlZQ1roC9CWlPk6Avf8EEnZNcAqPonwkG35x4n3ww/1THYAeQ==",
-      "dependencies": {
-        "node-domexception": "1.0.0",
-        "web-streams-polyfill": "4.0.0-beta.3"
-      },
-      "engines": {
-        "node": ">= 12.20"
-      }
-    },
-    "node_modules/humanize-ms": {
-      "version": "1.2.1",
-      "resolved": "https://registry.npmjs.org/humanize-ms/-/humanize-ms-1.2.1.tgz",
-      "integrity": "sha512-Fl70vYtsAFb/C06PTS9dZBo7ihau+Tu/DNCk/OyHhea07S+aeMWpFFkUaXRa8fI+ScZbEI8dfSxwY7gxZ9SAVQ==",
-      "dependencies": {
-        "ms": "^2.0.0"
-      }
-    },
-    "node_modules/is-any-array": {
-      "version": "2.0.1",
-      "resolved": "https://registry.npmjs.org/is-any-array/-/is-any-array-2.0.1.tgz",
-      "integrity": "sha512-UtilS7hLRu++wb/WBAw9bNuP1Eg04Ivn1vERJck8zJthEvXCBEBpGR/33u/xLKWEQf95803oalHrVDptcAvFdQ=="
-    },
-    "node_modules/is-buffer": {
-      "version": "1.1.6",
-      "resolved": "https://registry.npmjs.org/is-buffer/-/is-buffer-1.1.6.tgz",
-      "integrity": "sha512-NcdALwpXkTm5Zvvbk7owOUSvVvBKDgKP5/ewfXEznmQFfs4ZRmanOeKBTjRVjka3QFoN6XJ+9F3USqfHqTaU5w=="
-    },
-    "node_modules/js-tiktoken": {
-      "version": "1.0.7",
-      "resolved": "https://registry.npmjs.org/js-tiktoken/-/js-tiktoken-1.0.7.tgz",
-      "integrity": "sha512-biba8u/clw7iesNEWLOLwrNGoBP2lA+hTaBLs/D45pJdUPFXyxD6nhcDVtADChghv4GgyAiMKYMiRx7x6h7Biw==",
-      "dependencies": {
-        "base64-js": "^1.5.1"
-      }
-    },
-    "node_modules/js-yaml": {
-      "version": "4.1.0",
-      "resolved": "https://registry.npmjs.org/js-yaml/-/js-yaml-4.1.0.tgz",
-      "integrity": "sha512-wpxZs9NoxZaJESJGIZTyDEaYpl0FKSA+FB9aJiyemKhMwkxQg63h4T1KJgUGHpTqPDNRcmmYLugrRjJlBtWvRA==",
-      "dependencies": {
-        "argparse": "^2.0.1"
-      },
-      "bin": {
-        "js-yaml": "bin/js-yaml.js"
-      }
-    },
-    "node_modules/jsonpointer": {
-      "version": "5.0.1",
-      "resolved": "https://registry.npmjs.org/jsonpointer/-/jsonpointer-5.0.1.tgz",
-      "integrity": "sha512-p/nXbhSEcu3pZRdkW1OfJhpsVtW1gd4Wa1fnQc9YLiTfAjn0312eMKimbdIQzuZl9aa9xUGaRlP9T/CJE/ditQ==",
-      "engines": {
-        "node": ">=0.10.0"
-      }
-    },
-    "node_modules/langchain": {
-      "version": "0.0.165",
-      "resolved": "https://registry.npmjs.org/langchain/-/langchain-0.0.165.tgz",
-      "integrity": "sha512-CpbNpjwaE+9lzjdw+pZz0VgnRrFivEgr7CVp9dDaAb5JpaJAA4V2v6uQ9ZPN+TSqupTQ79HFn2sfyZVEl2EG7Q==",
-      "dependencies": {
-        "@anthropic-ai/sdk": "^0.6.2",
-        "ansi-styles": "^5.0.0",
-        "binary-extensions": "^2.2.0",
-        "camelcase": "6",
-        "decamelize": "^1.2.0",
-        "expr-eval": "^2.0.2",
-        "flat": "^5.0.2",
-        "js-tiktoken": "^1.0.7",
-        "js-yaml": "^4.1.0",
-        "jsonpointer": "^5.0.1",
-        "langchainhub": "~0.0.6",
-        "langsmith": "~0.0.31",
-        "ml-distance": "^4.0.0",
-        "object-hash": "^3.0.0",
-        "openai": "~4.4.0",
-        "openapi-types": "^12.1.3",
-        "p-queue": "^6.6.2",
-        "p-retry": "4",
-        "uuid": "^9.0.0",
-        "yaml": "^2.2.1",
-        "zod": "^3.22.3",
-        "zod-to-json-schema": "^3.20.4"
-      },
-      "engines": {
-        "node": ">=18"
-      },
-      "peerDependencies": {
-        "@aws-crypto/sha256-js": "^5.0.0",
-        "@aws-sdk/client-bedrock-runtime": "^3.422.0",
-        "@aws-sdk/client-dynamodb": "^3.310.0",
-        "@aws-sdk/client-kendra": "^3.352.0",
-        "@aws-sdk/client-lambda": "^3.310.0",
-        "@aws-sdk/client-s3": "^3.310.0",
-        "@aws-sdk/client-sagemaker-runtime": "^3.310.0",
-        "@aws-sdk/client-sfn": "^3.310.0",
-        "@aws-sdk/credential-provider-node": "^3.388.0",
-        "@azure/storage-blob": "^12.15.0",
-        "@clickhouse/client": "^0.0.14",
-        "@cloudflare/ai": "^1.0.12",
-        "@elastic/elasticsearch": "^8.4.0",
-        "@getmetal/metal-sdk": "*",
-        "@getzep/zep-js": "^0.7.0",
-        "@gomomento/sdk": "^1.23.0",
-        "@google-ai/generativelanguage": "^0.2.1",
-        "@google-cloud/storage": "^6.10.1",
-        "@huggingface/inference": "^1.5.1",
-        "@mozilla/readability": "*",
-        "@notionhq/client": "^2.2.10",
-        "@opensearch-project/opensearch": "*",
-        "@pinecone-database/pinecone": "^1.1.0",
-        "@planetscale/database": "^1.8.0",
-        "@qdrant/js-client-rest": "^1.2.0",
-        "@raycast/api": "^1.55.2",
-        "@smithy/eventstream-codec": "^2.0.5",
-        "@smithy/protocol-http": "^3.0.6",
-        "@smithy/signature-v4": "^2.0.10",
-        "@smithy/util-utf8": "^2.0.0",
-        "@supabase/postgrest-js": "^1.1.1",
-        "@supabase/supabase-js": "^2.10.0",
-        "@tensorflow-models/universal-sentence-encoder": "*",
-        "@tensorflow/tfjs-converter": "*",
-        "@tensorflow/tfjs-core": "*",
-        "@upstash/redis": "^1.20.6",
-        "@vercel/postgres": "^0.5.0",
-        "@writerai/writer-sdk": "^0.40.2",
-        "@xata.io/client": "^0.25.1",
-        "@xenova/transformers": "^2.5.4",
-        "@zilliz/milvus2-sdk-node": ">=2.2.7",
-        "apify-client": "^2.7.1",
-        "axios": "*",
-        "cassandra-driver": "^4.6.4",
-        "cheerio": "^1.0.0-rc.12",
-        "chromadb": "*",
-        "cohere-ai": ">=6.0.0",
-        "d3-dsv": "^2.0.0",
-        "epub2": "^3.0.1",
-        "faiss-node": "^0.3.0",
-        "fast-xml-parser": "^4.2.7",
-        "firebase-admin": "^11.9.0",
-        "google-auth-library": "^8.9.0",
-        "googleapis": "^126.0.1",
-        "hnswlib-node": "^1.4.2",
-        "html-to-text": "^9.0.5",
-        "ignore": "^5.2.0",
-        "ioredis": "^5.3.2",
-        "jsdom": "*",
-        "llmonitor": "*",
-        "lodash": "^4.17.21",
-        "mammoth": "*",
-        "mongodb": "^5.2.0",
-        "mysql2": "^3.3.3",
-        "neo4j-driver": "*",
-        "node-llama-cpp": "*",
-        "notion-to-md": "^3.1.0",
-        "pdf-parse": "1.1.1",
-        "peggy": "^3.0.2",
-        "pg": "^8.11.0",
-        "pg-copy-streams": "^6.0.5",
-        "pickleparser": "^0.1.0",
-        "playwright": "^1.32.1",
-        "portkey-ai": "^0.1.11",
-        "puppeteer": "^19.7.2",
-        "redis": "^4.6.4",
-        "replicate": "^0.18.0",
-        "sonix-speech-recognition": "^2.1.1",
-        "srt-parser-2": "^1.2.2",
-        "typeorm": "^0.3.12",
-        "typesense": "^1.5.3",
-        "usearch": "^1.1.1",
-        "vectordb": "^0.1.4",
-        "voy-search": "0.6.2",
-        "weaviate-ts-client": "^1.4.0",
-        "web-auth-library": "^1.0.3",
-        "youtube-transcript": "^1.0.6",
-        "youtubei.js": "^5.8.0"
-      },
-      "peerDependenciesMeta": {
-        "@aws-crypto/sha256-js": {
-          "optional": true
-        },
-        "@aws-sdk/client-bedrock-runtime": {
-          "optional": true
-        },
-        "@aws-sdk/client-dynamodb": {
-          "optional": true
-        },
-        "@aws-sdk/client-kendra": {
-          "optional": true
-        },
-        "@aws-sdk/client-lambda": {
-          "optional": true
-        },
-        "@aws-sdk/client-s3": {
-          "optional": true
-        },
-        "@aws-sdk/client-sagemaker-runtime": {
-          "optional": true
-        },
-        "@aws-sdk/client-sfn": {
-          "optional": true
-        },
-        "@aws-sdk/credential-provider-node": {
-          "optional": true
-        },
-        "@azure/storage-blob": {
-          "optional": true
-        },
-        "@clickhouse/client": {
-          "optional": true
-        },
-        "@cloudflare/ai": {
-          "optional": true
-        },
-        "@elastic/elasticsearch": {
-          "optional": true
-        },
-        "@getmetal/metal-sdk": {
-          "optional": true
-        },
-        "@getzep/zep-js": {
-          "optional": true
-        },
-        "@gomomento/sdk": {
-          "optional": true
-        },
-        "@google-ai/generativelanguage": {
-          "optional": true
-        },
-        "@google-cloud/storage": {
-          "optional": true
-        },
-        "@huggingface/inference": {
-          "optional": true
-        },
-        "@mozilla/readability": {
-          "optional": true
-        },
-        "@notionhq/client": {
-          "optional": true
-        },
-        "@opensearch-project/opensearch": {
-          "optional": true
-        },
-        "@pinecone-database/pinecone": {
-          "optional": true
-        },
-        "@planetscale/database": {
-          "optional": true
-        },
-        "@qdrant/js-client-rest": {
-          "optional": true
-        },
-        "@raycast/api": {
-          "optional": true
-        },
-        "@smithy/eventstream-codec": {
-          "optional": true
-        },
-        "@smithy/protocol-http": {
-          "optional": true
-        },
-        "@smithy/signature-v4": {
-          "optional": true
-        },
-        "@smithy/util-utf8": {
-          "optional": true
-        },
-        "@supabase/postgrest-js": {
-          "optional": true
-        },
-        "@supabase/supabase-js": {
-          "optional": true
-        },
-        "@tensorflow-models/universal-sentence-encoder": {
-          "optional": true
-        },
-        "@tensorflow/tfjs-converter": {
-          "optional": true
-        },
-        "@tensorflow/tfjs-core": {
-          "optional": true
-        },
-        "@upstash/redis": {
-          "optional": true
-        },
-        "@vercel/postgres": {
-          "optional": true
-        },
-        "@writerai/writer-sdk": {
-          "optional": true
-        },
-        "@xata.io/client": {
-          "optional": true
-        },
-        "@xenova/transformers": {
-          "optional": true
-        },
-        "@zilliz/milvus2-sdk-node": {
-          "optional": true
-        },
-        "apify-client": {
-          "optional": true
-        },
-        "axios": {
-          "optional": true
-        },
-        "cassandra-driver": {
-          "optional": true
-        },
-        "cheerio": {
-          "optional": true
-        },
-        "chromadb": {
-          "optional": true
-        },
-        "cohere-ai": {
-          "optional": true
-        },
-        "d3-dsv": {
-          "optional": true
-        },
-        "epub2": {
-          "optional": true
-        },
-        "faiss-node": {
-          "optional": true
-        },
-        "fast-xml-parser": {
-          "optional": true
-        },
-        "firebase-admin": {
-          "optional": true
-        },
-        "google-auth-library": {
-          "optional": true
-        },
-        "googleapis": {
-          "optional": true
-        },
-        "hnswlib-node": {
-          "optional": true
-        },
-        "html-to-text": {
-          "optional": true
-        },
-        "ignore": {
-          "optional": true
-        },
-        "ioredis": {
-          "optional": true
-        },
-        "jsdom": {
-          "optional": true
-        },
-        "llmonitor": {
-          "optional": true
-        },
-        "lodash": {
-          "optional": true
-        },
-        "mammoth": {
-          "optional": true
-        },
-        "mongodb": {
-          "optional": true
-        },
-        "mysql2": {
-          "optional": true
-        },
-        "neo4j-driver": {
-          "optional": true
-        },
-        "node-llama-cpp": {
-          "optional": true
-        },
-        "notion-to-md": {
-          "optional": true
-        },
-        "pdf-parse": {
-          "optional": true
-        },
-        "peggy": {
-          "optional": true
-        },
-        "pg": {
-          "optional": true
-        },
-        "pg-copy-streams": {
-          "optional": true
-        },
-        "pickleparser": {
-          "optional": true
-        },
-        "playwright": {
-          "optional": true
-        },
-        "portkey-ai": {
-          "optional": true
-        },
-        "puppeteer": {
-          "optional": true
-        },
-        "redis": {
-          "optional": true
-        },
-        "replicate": {
-          "optional": true
-        },
-        "sonix-speech-recognition": {
-          "optional": true
-        },
-        "srt-parser-2": {
-          "optional": true
-        },
-        "typeorm": {
-          "optional": true
-        },
-        "typesense": {
-          "optional": true
-        },
-        "usearch": {
-          "optional": true
-        },
-        "vectordb": {
-          "optional": true
-        },
-        "voy-search": {
-          "optional": true
-        },
-        "weaviate-ts-client": {
-          "optional": true
-        },
-        "web-auth-library": {
-          "optional": true
-        },
-        "youtube-transcript": {
-          "optional": true
-        },
-        "youtubei.js": {
-          "optional": true
-        }
-      }
-    },
-    "node_modules/langchainhub": {
-      "version": "0.0.6",
-      "resolved": "https://registry.npmjs.org/langchainhub/-/langchainhub-0.0.6.tgz",
-      "integrity": "sha512-SW6105T+YP1cTe0yMf//7kyshCgvCTyFBMTgH2H3s9rTAR4e+78DA/BBrUL/Mt4Q5eMWui7iGuAYb3pgGsdQ9w=="
-    },
-    "node_modules/langsmith": {
-      "version": "0.0.42",
-      "resolved": "https://registry.npmjs.org/langsmith/-/langsmith-0.0.42.tgz",
-      "integrity": "sha512-sFuN+e7E+pPBIRaRgFqZh/BRBWNHTZNAwi6uj4kydQawooCZYoJmM5snOkiQrhVSvAhgu6xFhLvmfvkPcKzD7w==",
-      "dependencies": {
-        "@types/uuid": "^9.0.1",
-        "commander": "^10.0.1",
-        "p-queue": "^6.6.2",
-        "p-retry": "4",
-        "uuid": "^9.0.0"
-      },
-      "bin": {
-        "langsmith": "dist/cli/main.cjs"
-      }
-    },
-    "node_modules/md5": {
-      "version": "2.3.0",
-      "resolved": "https://registry.npmjs.org/md5/-/md5-2.3.0.tgz",
-      "integrity": "sha512-T1GITYmFaKuO91vxyoQMFETst+O71VUPEU3ze5GNzDm0OWdP8v1ziTaAEPUr/3kLsY3Sftgz242A1SetQiDL7g==",
-      "dependencies": {
-        "charenc": "0.0.2",
-        "crypt": "0.0.2",
-        "is-buffer": "~1.1.6"
-      }
-    },
-    "node_modules/mime-db": {
-      "version": "1.52.0",
-      "resolved": "https://registry.npmjs.org/mime-db/-/mime-db-1.52.0.tgz",
-      "integrity": "sha512-sPU4uV7dYlvtWJxwwxHD0PuihVNiE7TyAbQ5SWxDCB9mUYvOgroQOwYQQOKPJ8CIbE+1ETVlOoK1UC2nU3gYvg==",
-      "engines": {
-        "node": ">= 0.6"
-      }
-    },
-    "node_modules/mime-types": {
-      "version": "2.1.35",
-      "resolved": "https://registry.npmjs.org/mime-types/-/mime-types-2.1.35.tgz",
-      "integrity": "sha512-ZDY+bPm5zTTF+YpCrAU9nK0UgICYPT0QtT1NZWFv4s++TNkcgVaT0g6+4R2uI4MjQjzysHB1zxuWL50hzaeXiw==",
-      "dependencies": {
-        "mime-db": "1.52.0"
-      },
-      "engines": {
-        "node": ">= 0.6"
-      }
-    },
-    "node_modules/ml-array-mean": {
-      "version": "1.1.6",
-      "resolved": "https://registry.npmjs.org/ml-array-mean/-/ml-array-mean-1.1.6.tgz",
-      "integrity": "sha512-MIdf7Zc8HznwIisyiJGRH9tRigg3Yf4FldW8DxKxpCCv/g5CafTw0RRu51nojVEOXuCQC7DRVVu5c7XXO/5joQ==",
-      "dependencies": {
-        "ml-array-sum": "^1.1.6"
-      }
-    },
-    "node_modules/ml-array-sum": {
-      "version": "1.1.6",
-      "resolved": "https://registry.npmjs.org/ml-array-sum/-/ml-array-sum-1.1.6.tgz",
-      "integrity": "sha512-29mAh2GwH7ZmiRnup4UyibQZB9+ZLyMShvt4cH4eTK+cL2oEMIZFnSyB3SS8MlsTh6q/w/yh48KmqLxmovN4Dw==",
-      "dependencies": {
-        "is-any-array": "^2.0.0"
-      }
-    },
-    "node_modules/ml-distance": {
-      "version": "4.0.1",
-      "resolved": "https://registry.npmjs.org/ml-distance/-/ml-distance-4.0.1.tgz",
-      "integrity": "sha512-feZ5ziXs01zhyFUUUeZV5hwc0f5JW0Sh0ckU1koZe/wdVkJdGxcP06KNQuF0WBTj8FttQUzcvQcpcrOp/XrlEw==",
-      "dependencies": {
-        "ml-array-mean": "^1.1.6",
-        "ml-distance-euclidean": "^2.0.0",
-        "ml-tree-similarity": "^1.0.0"
-      }
-    },
-    "node_modules/ml-distance-euclidean": {
-      "version": "2.0.0",
-      "resolved": "https://registry.npmjs.org/ml-distance-euclidean/-/ml-distance-euclidean-2.0.0.tgz",
-      "integrity": "sha512-yC9/2o8QF0A3m/0IXqCTXCzz2pNEzvmcE/9HFKOZGnTjatvBbsn4lWYJkxENkA4Ug2fnYl7PXQxnPi21sgMy/Q=="
-    },
-    "node_modules/ml-tree-similarity": {
-      "version": "1.0.0",
-      "resolved": "https://registry.npmjs.org/ml-tree-similarity/-/ml-tree-similarity-1.0.0.tgz",
-      "integrity": "sha512-XJUyYqjSuUQkNQHMscr6tcjldsOoAekxADTplt40QKfwW6nd++1wHWV9AArl0Zvw/TIHgNaZZNvr8QGvE8wLRg==",
-      "dependencies": {
-        "binary-search": "^1.3.5",
-        "num-sort": "^2.0.0"
-      }
-    },
-    "node_modules/ms": {
-      "version": "2.1.3",
-      "resolved": "https://registry.npmjs.org/ms/-/ms-2.1.3.tgz",
-      "integrity": "sha512-6FlzubTLZG3J2a/NVCAleEhjzq5oxgHyaCU9yYXvcLsvoVaHJq/s5xXI6/XXP6tz7R9xAOtHnSO/tXtF3WRTlA=="
-    },
-    "node_modules/node-domexception": {
-      "version": "1.0.0",
-      "resolved": "https://registry.npmjs.org/node-domexception/-/node-domexception-1.0.0.tgz",
-      "integrity": "sha512-/jKZoMpw0F8GRwl4/eLROPA3cfcXtLApP0QzLmUT/HuPCZWyB7IY9ZrMeKw2O/nFIqPQB3PVM9aYm0F312AXDQ==",
-      "funding": [
-        {
-          "type": "github",
-          "url": "https://github.com/sponsors/jimmywarting"
-        },
-        {
-          "type": "github",
-          "url": "https://paypal.me/jimmywarting"
-        }
-      ],
-      "engines": {
-        "node": ">=10.5.0"
-      }
-    },
-    "node_modules/node-fetch": {
-      "version": "2.7.0",
-      "resolved": "https://registry.npmjs.org/node-fetch/-/node-fetch-2.7.0.tgz",
-      "integrity": "sha512-c4FRfUm/dbcWZ7U+1Wq0AwCyFL+3nt2bEw05wfxSz+DWpWsitgmSgYmy2dQdWyKC1694ELPqMs/YzUSNozLt8A==",
-      "dependencies": {
-        "whatwg-url": "^5.0.0"
-      },
-      "engines": {
-        "node": "4.x || >=6.0.0"
-      },
-      "peerDependencies": {
-        "encoding": "^0.1.0"
-      },
-      "peerDependenciesMeta": {
-        "encoding": {
-          "optional": true
-        }
-      }
-    },
-    "node_modules/num-sort": {
-      "version": "2.1.0",
-      "resolved": "https://registry.npmjs.org/num-sort/-/num-sort-2.1.0.tgz",
-      "integrity": "sha512-1MQz1Ed8z2yckoBeSfkQHHO9K1yDRxxtotKSJ9yvcTUUxSvfvzEq5GwBrjjHEpMlq/k5gvXdmJ1SbYxWtpNoVg==",
-      "engines": {
-        "node": ">=8"
-      },
-      "funding": {
-        "url": "https://github.com/sponsors/sindresorhus"
-      }
-    },
-    "node_modules/object-hash": {
-      "version": "3.0.0",
-      "resolved": "https://registry.npmjs.org/object-hash/-/object-hash-3.0.0.tgz",
-      "integrity": "sha512-RSn9F68PjH9HqtltsSnqYC1XXoWe9Bju5+213R98cNGttag9q9yAOTzdbsqvIa7aNm5WffBZFpWYr2aWrklWAw==",
-      "engines": {
-        "node": ">= 6"
-      }
-    },
-    "node_modules/openai": {
-      "version": "4.4.0",
-      "resolved": "https://registry.npmjs.org/openai/-/openai-4.4.0.tgz",
-      "integrity": "sha512-JN0t628Kh95T0IrXl0HdBqnlJg+4Vq0Bnh55tio+dfCnyzHvMLiWyCM9m726MAJD2YkDU4/8RQB6rNbEq9ct2w==",
-      "dependencies": {
-        "@types/node": "^18.11.18",
-        "@types/node-fetch": "^2.6.4",
-        "abort-controller": "^3.0.0",
-        "agentkeepalive": "^4.2.1",
-        "digest-fetch": "^1.3.0",
-        "form-data-encoder": "1.7.2",
-        "formdata-node": "^4.3.2",
-        "node-fetch": "^2.6.7"
-      },
-      "bin": {
-        "openai": "bin/cli"
-      }
-    },
-    "node_modules/openapi-types": {
-      "version": "12.1.3",
-      "resolved": "https://registry.npmjs.org/openapi-types/-/openapi-types-12.1.3.tgz",
-      "integrity": "sha512-N4YtSYJqghVu4iek2ZUvcN/0aqH1kRDuNqzcycDxhOUpg7GdvLa2F3DgS6yBNhInhv2r/6I0Flkn7CqL8+nIcw=="
-    },
-    "node_modules/p-finally": {
-      "version": "1.0.0",
-      "resolved": "https://registry.npmjs.org/p-finally/-/p-finally-1.0.0.tgz",
-      "integrity": "sha512-LICb2p9CB7FS+0eR1oqWnHhp0FljGLZCWBE9aix0Uye9W8LTQPwMTYVGWQWIw9RdQiDg4+epXQODwIYJtSJaow==",
-      "engines": {
-        "node": ">=4"
-      }
-    },
-    "node_modules/p-queue": {
-      "version": "6.6.2",
-      "resolved": "https://registry.npmjs.org/p-queue/-/p-queue-6.6.2.tgz",
-      "integrity": "sha512-RwFpb72c/BhQLEXIZ5K2e+AhgNVmIejGlTgiB9MzZ0e93GRvqZ7uSi0dvRF7/XIXDeNkra2fNHBxTyPDGySpjQ==",
-      "dependencies": {
-        "eventemitter3": "^4.0.4",
-        "p-timeout": "^3.2.0"
-      },
-      "engines": {
-        "node": ">=8"
-      },
-      "funding": {
-        "url": "https://github.com/sponsors/sindresorhus"
-      }
-    },
-    "node_modules/p-retry": {
-      "version": "4.6.2",
-      "resolved": "https://registry.npmjs.org/p-retry/-/p-retry-4.6.2.tgz",
-      "integrity": "sha512-312Id396EbJdvRONlngUx0NydfrIQ5lsYu0znKVUzVvArzEIt08V1qhtyESbGVd1FGX7UKtiFp5uwKZdM8wIuQ==",
-      "dependencies": {
-        "@types/retry": "0.12.0",
-        "retry": "^0.13.1"
-      },
-      "engines": {
-        "node": ">=8"
-      }
-    },
-    "node_modules/p-timeout": {
-      "version": "3.2.0",
-      "resolved": "https://registry.npmjs.org/p-timeout/-/p-timeout-3.2.0.tgz",
-      "integrity": "sha512-rhIwUycgwwKcP9yTOOFK/AKsAopjjCakVqLHePO3CC6Mir1Z99xT+R63jZxAT5lFZLa2inS5h+ZS2GvR99/FBg==",
-      "dependencies": {
-        "p-finally": "^1.0.0"
-      },
-      "engines": {
-        "node": ">=8"
-      }
-    },
-    "node_modules/retry": {
-      "version": "0.13.1",
-      "resolved": "https://registry.npmjs.org/retry/-/retry-0.13.1.tgz",
-      "integrity": "sha512-XQBQ3I8W1Cge0Seh+6gjj03LbmRFWuoszgK9ooCpwYIrhhoO80pfq4cUkU5DkknwfOfFteRwlZ56PYOGYyFWdg==",
-      "engines": {
-        "node": ">= 4"
-      }
-    },
-    "node_modules/tr46": {
-      "version": "0.0.3",
-      "resolved": "https://registry.npmjs.org/tr46/-/tr46-0.0.3.tgz",
-      "integrity": "sha512-N3WMsuqV66lT30CrXNbEjx4GEwlow3v6rr4mCcv6prnfwhS01rkgyFdjPNBYd9br7LpXV1+Emh01fHnq2Gdgrw=="
-    },
-    "node_modules/typescript": {
-      "version": "5.2.2",
-      "resolved": "https://registry.npmjs.org/typescript/-/typescript-5.2.2.tgz",
-      "integrity": "sha512-mI4WrpHsbCIcwT9cF4FZvr80QUeKvsUsUvKDoR+X/7XHQH98xYD8YHZg7ANtz2GtZt/CBq2QJ0thkGJMHfqc1w==",
-      "dev": true,
-      "bin": {
-        "tsc": "bin/tsc",
-        "tsserver": "bin/tsserver"
-      },
-      "engines": {
-        "node": ">=14.17"
-      }
-    },
-    "node_modules/uuid": {
-      "version": "9.0.1",
-      "resolved": "https://registry.npmjs.org/uuid/-/uuid-9.0.1.tgz",
-      "integrity": "sha512-b+1eJOlsR9K8HJpow9Ok3fiWOWSIcIzXodvv0rQjVoOVNpWMpxf1wZNpt4y9h10odCNrqnYp1OBzRktckBe3sA==",
-      "funding": [
-        "https://github.com/sponsors/broofa",
-        "https://github.com/sponsors/ctavan"
-      ],
-      "bin": {
-        "uuid": "dist/bin/uuid"
-      }
-    },
-    "node_modules/web-streams-polyfill": {
-      "version": "4.0.0-beta.3",
-      "resolved": "https://registry.npmjs.org/web-streams-polyfill/-/web-streams-polyfill-4.0.0-beta.3.tgz",
-      "integrity": "sha512-QW95TCTaHmsYfHDybGMwO5IJIM93I/6vTRk+daHTWFPhwh+C8Cg7j7XyKrwrj8Ib6vYXe0ocYNrmzY4xAAN6ug==",
-      "engines": {
-        "node": ">= 14"
-      }
-    },
-    "node_modules/webidl-conversions": {
-      "version": "3.0.1",
-      "resolved": "https://registry.npmjs.org/webidl-conversions/-/webidl-conversions-3.0.1.tgz",
-      "integrity": "sha512-2JAn3z8AR6rjK8Sm8orRC0h/bcl/DqL7tRPdGZ4I1CjdF+EaMLmYxBHyXuKL849eucPFhvBoxMsflfOb8kxaeQ=="
-    },
-    "node_modules/whatwg-url": {
-      "version": "5.0.0",
-      "resolved": "https://registry.npmjs.org/whatwg-url/-/whatwg-url-5.0.0.tgz",
-      "integrity": "sha512-saE57nupxk6v3HY35+jzBwYa0rKSy0XR8JSxZPwgLr7ys0IBzhGviA1/TUGJLmSVqs8pb9AnvICXEuOHLprYTw==",
-      "dependencies": {
-        "tr46": "~0.0.3",
-        "webidl-conversions": "^3.0.0"
-      }
-    },
-    "node_modules/yaml": {
-      "version": "2.3.2",
-      "resolved": "https://registry.npmjs.org/yaml/-/yaml-2.3.2.tgz",
-      "integrity": "sha512-N/lyzTPaJasoDmfV7YTrYCI0G/3ivm/9wdG0aHuheKowWQwGTsK0Eoiw6utmzAnI6pkJa0DUVygvp3spqqEKXg==",
-      "engines": {
-        "node": ">= 14"
-      }
-    },
-    "node_modules/zod": {
-      "version": "3.22.4",
-      "resolved": "https://registry.npmjs.org/zod/-/zod-3.22.4.tgz",
-      "integrity": "sha512-iC+8Io04lddc+mVqQ9AZ7OQ2MrUKGN+oIQyq1vemgt46jwCwLfhq7/pwnBnNXXXZb8VTVLKwp9EDkx+ryxIWmg==",
-      "funding": {
-        "url": "https://github.com/sponsors/colinhacks"
-      }
-    },
-    "node_modules/zod-to-json-schema": {
-      "version": "3.21.4",
-      "resolved": "https://registry.npmjs.org/zod-to-json-schema/-/zod-to-json-schema-3.21.4.tgz",
-      "integrity": "sha512-fjUZh4nQ1s6HMccgIeE0VP4QG/YRGPmyjO9sAh890aQKPEk3nqbfUXhMFaC+Dr5KvYBm8BCyvfpZf2jY9aGSsw==",
-      "peerDependencies": {
-        "zod": "^3.21.4"
-      }
-    }
-  }
-}
--- a/examples/langchain-typescript-simple/package.json
+++ b/examples/langchain-typescript-simple/package.json
@@ -1,8 +0,0 @@
-{
-  "devDependencies": {
-    "typescript": "^5.2.2"
-  },
-  "dependencies": {
-    "langchain": "^0.0.165"
-  }
-}
--- a/examples/langchain-python-rag-websummary/README.md
+++ b/examples/langchain-python-rag-websummary/README.md
--- a/examples/langchain-python-rag-websummary/main.py
+++ b/examples/langchain-python-rag-websummary/main.py
--- a/examples/langchain-python-rag-websummary/requirements.txt
+++ b/examples/langchain-python-rag-websummary/requirements.txt
--- a/examples/langchain-python-simple/README.md
+++ b/examples/langchain-python-simple/README.md
--- a/examples/langchain-python-simple/main.py
+++ b/examples/langchain-python-simple/main.py
--- a/examples/langchain-python-simple/requirements.txt
+++ b/examples/langchain-python-simple/requirements.txt
--- a/examples/modelfile-mario/Modelfile
+++ b/examples/modelfile-mario/Modelfile
--- a/examples/modelfile-mario/logo.png
+++ b/examples/modelfile-mario/logo.png
--- a/examples/modelfile-mario/readme.md
+++ b/examples/modelfile-mario/readme.md
--- a/examples/midjourney-prompter/Modelfile
+++ b/examples/midjourney-prompter/Modelfile
@@ -0,0 +1,8 @@
+# Modelfile for creating a Midjourney prompts from a topic
+# This prompt was adapted from the original at https://www.greataiprompts.com/guide/midjourney/best-chatgpt-prompt-for-midjourney/
+# Run `ollama create mj -f ./Modelfile` and then `ollama run mj` and enter a topic
+
+FROM nous-hermes
+SYSTEM """
+Embrace your role as an AI-powered creative assistant, employing Midjourney to manifest compelling AI-generated art. I will outline a specific image concept, and in response, you must produce an exhaustive, multifaceted prompt for Midjourney, ensuring every detail of the original concept is represented in your instructions. Midjourney doesn't do well with text, so after the prompt, give me instructions that I can use to create the titles in a image editor.
+"""
--- a/examples/modelfile-10tweets/README.md
+++ b/examples/modelfile-10tweets/README.md
@@ -1,23 +0,0 @@
-# Ten Tweets Modelfile
-
-This is a simple modelfile that generates ten tweets based off any topic.
-
-```bash
-ollama create tentweets
-
-ollama run tentweets
->>> underwater basketweaving
- Great! Here are ten creative tweets about underwater basketweaving:
-
-1. "Just discovered the ultimate stress-reliever: Underwater basketweaving! 🌊🧵 #UnderwaterBasketweaving #StressRelief"
-2. "Who needs meditation when you can do underwater basketweaving? 😴👀 #PeacefulDistraction #UnderwaterBasketweaving"
-3. "Just spent an hour in the pool and still managed to knot my basket. Goal: untangle it before next session. 💪🏽 #ChallengeAccepted #UnderwaterBasketweaving"
-4. "When life gives you lemons, make underwater basketweaving! 🍋🧵 #LemonadeLife #UnderwaterBasketweaving"
-5. "Just realized my underwater basketweaving skills could come in handy during a zombie apocalypse. 😂🧡 #SurvivalTips #UnderwaterBasketweaving"
-6. "I'm not lazy, I'm just conserving energy for my next underwater basketweaving session. 😴💤 #LazyDay #UnderwaterBasketweaving"
-7. "Just found my inner peace while doing underwater basketweaving. It's like meditation, but with knots! 🙏🧵 #Mindfulness #UnderwaterBasketweaving"
-8. "Why study for exams when you can do underwater basketweaving and forget all your worries? 😜🧵 #ProcrastinationStation #UnderwaterBasketweaving"
-9. "Just had to cut my underwater basketweaving session short due to a sudden urge to breathe. 🤯🌊 #AquaticAdventures #UnderwaterBasketweaving"
-10. "I'm not sure what's more impressive: my underwater basketweaving skills or the fact that I didn't drown trying to make this tweet. 😅🧵 #Accomplishment 
-#UnderwaterBasketweaving"
-```
--- a/examples/modelfile-devopsengineer/README.md
+++ b/examples/modelfile-devopsengineer/README.md
@@ -1,40 +0,0 @@
-# DevOps Engineer Modelfile
-
-This is a simple modelfile that takes will respond to any questions that a senior DevOps engineer should know how to solve.
-
-````text
-ollama create devopsengineer
-
-ollama run devopsengineer
->>> how can i proxy all connections from inside the office through a single server that caches large files. Provide a config file for nginx to achieve this?
-
-Here is an example configuration file for Nginx that can be used to proxy all connections from inside the office network to a centralized cache server:
-```perl
-http {
-    upstream cache_server {
-        server <cache_server_ip>;
-        server <cache_server_ip>;
-        server <cache_server_ip>;
-    }
-    
-    server {
-        listen 80;
-        server_name localhost;
-        
-        location / {
-            proxy_pass http://cache_server;
-            proxy_set_header Host $host;
-            proxy_set_header X-Real-IP $remote_addr;
-            proxy_set_header X-Forwarded-For $proxy_add_x_forwarded_for;
-        }
-    }
-}
-```
-In this configuration, we define an `upstream` block that lists the IP addresses of the cache servers. We then define a server block that listens on port 80 and routes all 
-traffic to the `/` location to the cache servers using the `proxy_pass` directive. The `proxy_set_header` directives are used to preserve the source IP address of the client
-request when forwarding it to the cache server.
-
-To use this configuration, you would need to replace the placeholder `<cache_server_ip>` with the actual IP addresses of your cache servers. You would also need to make sure
-that the cache servers are configured to accept incoming connections from the Nginx server and handle requests for files.
-
-````
--- a/examples/modelfile-midjourney/Modelfile
+++ b/examples/modelfile-midjourney/Modelfile
@@ -1,11 +0,0 @@
-# Modelfile for creating a Midjourney prompts from a topic
-# This prompt was adapted from the original at https://www.greataiprompts.com/guide/midjourney/best-chatgpt-prompt-for-midjourney/
-# Run `ollama create mj -f ./Modelfile` and then `ollama run mj` and enter a topic
-
-FROM zephyr
-PARAMETER temperature 0.8
-PARAMETER top_k 500
-PARAMETER top_p 0.9
-SYSTEM """
-Embrace your role as a creative illustrator. Based on a concept provided, you must produce a single paragraph with a multifaceted description of an image, ensuring significant details of the concept and more is represented in your instructions. You do not need to write complete sentences but rather short concepts with the following information: the level of detail that should be represented, an artistic style and maybe a specific name of a painter or illustrator, the ideal color pallete, lighting, mood, perspective, the setting, time of day, weather, the season, the time period, location, materials, the textures, patterns, lines, brushstrokes, techniques, the medium, the genre, the rendering style. Don't include everything and keep the description length under 250 words. 
-"""
--- a/examples/modelfile-midjourney/README.md
+++ b/examples/modelfile-midjourney/README.md
@@ -1,11 +0,0 @@
-# Midjourney Prompt Generator Modelfile
-
-This simple modelfile will help create a prompt to feed to Midjourney.
-
-```text
-ollama create midjourney
-
-ollama run midjourney
->>> a sports car in the mountains. 
-A sleek, high-performance automobile cuts through a serpentine mountain landscape. The concept is a classic illustration of speed and power, depicted in the style of pop art by Andy Warhol. The color palette is dominated by bold, primary hues of red, blue, and yellow, with striking accent colors of white, black, and metallic shades. The lighting is bright and focused, casting sharp shadows on the rugged terrain. A sense of excitement and anticipation permeates throughout the scene, as the car navigates a treacherous course through the winding road. The perspective is low, allowing for a full view of the vehicle's sleek lines and intricate details. The setting takes place in the afternoon during a sunny day in autumn, as evidenced by the vibrant foliage on the mountainside. The time period is modern, with nods to classic car design. The materials are primarily digital, allowing for smooth curves and sharp contrasts. The textures are sleek and polished, with meticulously detailed lines and brushstrokes that accentuate the car's aerodynamic design. The patterns consist of geometric shapes and bold stripes, adding to the car's dynamic appeal. The genre is modern realism, with a focus on precision and detail. The rendering style is highly technical, capturing the nuances and subtleties of the vehicle and its surroundings in breathtaking detail.
-```
--- a/examples/modelfile-recipemaker/README.md
+++ b/examples/modelfile-recipemaker/README.md
@@ -1,20 +0,0 @@
-# Recipe Maker Modelfile 
-
-Simple modelfile to generate a recipe from a short list of ingredients.
-
-```
-ollama create recipemaker
-
-ollama run recipemaker
->>> chilli pepper, white chocolate, kale
- Ingredients:
- 1 small chili pepper
- 4 squares of white chocolate
- handful of kale leaves
-
-Instructions:
-1. In a blender or food processor, puree the chilies and white chocolate until smooth.
-2. Add the chopped kale leaves to the blender and pulse until well combined.
-3. Serve immediately as a dip for crackers or use it as an ingredient in your favorite recipe. The mixture of spicy chili pepper with sweet white chocolate and nutritious 
-kale will make your taste buds dance with delight!
-```
--- a/examples/langchain-python-rag-privategpt/.gitignore
+++ b/examples/langchain-python-rag-privategpt/.gitignore
--- a/examples/langchain-python-rag-privategpt/LICENSE
+++ b/examples/langchain-python-rag-privategpt/LICENSE
--- a/examples/langchain-python-rag-privategpt/README.md
+++ b/examples/langchain-python-rag-privategpt/README.md
--- a/examples/langchain-python-rag-privategpt/constants.py
+++ b/examples/langchain-python-rag-privategpt/constants.py
--- a/examples/langchain-python-rag-privategpt/ingest.py
+++ b/examples/langchain-python-rag-privategpt/ingest.py
--- a/examples/langchain-python-rag-privategpt/poetry.lock
+++ b/examples/langchain-python-rag-privategpt/poetry.lock
--- a/examples/langchain-python-rag-privategpt/privateGPT.py
+++ b/examples/langchain-python-rag-privategpt/privateGPT.py
--- a/examples/langchain-python-rag-privategpt/pyproject.toml
+++ b/examples/langchain-python-rag-privategpt/pyproject.toml
--- a/examples/langchain-python-rag-privategpt/requirements.txt
+++ b/examples/langchain-python-rag-privategpt/requirements.txt
--- a/examples/python-simplegenerate/client.py
+++ b/examples/python-simplegenerate/client.py
--- a/examples/modelfile-recipemaker/Modelfile
+++ b/examples/modelfile-recipemaker/Modelfile
--- a/examples/modelfile-sentiments/Modelfile
+++ b/examples/modelfile-sentiments/Modelfile
--- a/examples/modelfile-sentiments/Readme.md
+++ b/examples/modelfile-sentiments/Readme.md
--- a/examples/modelfile-tweetwriter/Modelfile
+++ b/examples/modelfile-tweetwriter/Modelfile
--- a/examples/typescript-mentors/.gitignore
+++ b/examples/typescript-mentors/.gitignore
@@ -1,2 +0,0 @@
-node_modules
-package-lock.json
--- a/examples/typescript-mentors/README.md
+++ b/examples/typescript-mentors/README.md
@@ -1,21 +0,0 @@
-# Ask the Mentors
-
-This example demonstrates how one would create a set of 'mentors' you can have a conversation with. The mentors are generated using the `character-generator.ts` file. This will use **Stable Beluga 70b** to create a bio and list of verbal ticks and common phrases used by each person. Then `mentors.ts` will take a question, and choose three of the 'mentors' and start a conversation with them. Occasionally, they will talk to each other, and other times they will just deliver a set of monologues. It's fun to see what they do and say.
-
-## Usage
-
-```bash
-ts-node ./character-generator.ts "Lorne Greene"
-```
-
-This will create `lornegreene/Modelfile`. Now you can create a model with this command:
-
-```bash
-ollama create lornegreene -f lornegreene/Modelfile
-```
-
-If you want to add your own mentors, you will have to update the code to look at your namespace instead of **mattw**. Also set the list of mentors to include yours.
-
-```bash
-ts-node ./mentors.ts "What is a Jackalope?"
-```
--- a/examples/typescript-mentors/character-generator.ts
+++ b/examples/typescript-mentors/character-generator.ts
@@ -1,26 +0,0 @@
-import { Ollama } from 'ollama-node'
-import fs from 'fs';
-import path from 'path';
-
-async function characterGenerator() {
-  const character = process.argv[2];
-  console.log(`You are creating a character for ${character}.`);
-  const foldername = character.replace(/\s/g, '').toLowerCase();
-  const directory = path.join(__dirname, foldername);
-  if (!fs.existsSync(directory)) {
-    fs.mkdirSync(directory, { recursive: true });
-  }
-
-  const ollama = new Ollama();
-  ollama.setModel("stablebeluga2:70b-q4_K_M");
-  const bio = await ollama.generate(`create a bio of ${character} in a single long paragraph. Instead of saying '${character} is...' or '${character} was...' use language like 'You are...' or 'You were...'. Then create a paragraph describing the speaking mannerisms and style of ${character}. Don't include anything about how ${character} looked or what they sounded like, just focus on the words they said. Instead of saying '${character} would say...' use language like 'You should say...'. If you use quotes, always use single quotes instead of double quotes. If there are any specific words or phrases you used a lot, show how you used them. `);
-
-  const thecontents = `FROM llama2\nSYSTEM """\n${bio.response.replace(/(\r\n|\n|\r)/gm, " ").replace('would', 'should')} All answers to questions should be related back to what you are most known for.\n"""`;
-
-  fs.writeFile(path.join(directory, 'Modelfile'), thecontents, (err: any) => {
-    if (err) throw err;
-    console.log('The file has been saved!');
-  });
-}
-
-characterGenerator();
--- a/examples/typescript-mentors/mentors.ts
+++ b/examples/typescript-mentors/mentors.ts
@@ -1,59 +0,0 @@
-import { Ollama } from 'ollama-node';
-
-const mentorCount = 3;
-const ollama = new Ollama();
-
-function getMentors(): string[] {
-  const mentors = ['Gary Vaynerchuk', 'Kanye West', 'Martha Stewart', 'Neil deGrasse Tyson', 'Owen Wilson', 'Ronald Reagan', 'Donald Trump', 'Barack Obama', 'Jeff Bezos'];
-  const chosenMentors: string[] = [];
-  for (let i = 0; i < mentorCount; i++) {
-    const mentor = mentors[Math.floor(Math.random() * mentors.length)];
-    chosenMentors.push(mentor);
-    mentors.splice(mentors.indexOf(mentor), 1);
-  }
-  return chosenMentors;
-}
-
-function getMentorFileName(mentor: string): string {
-  const model = mentor.toLowerCase().replace(/\s/g, '');
-  return `mattw/${model}`;
-}
-
-async function getSystemPrompt(mentor: string, isLast: boolean, question: string): Promise<string> {
-  ollama.setModel(getMentorFileName(mentor));
-  const info = await ollama.showModelInfo()
-  let SystemPrompt = info.system || '';
-  SystemPrompt += ` You should continue the conversation as if you were ${mentor} and acknowledge the people before you in the conversation. You should adopt their mannerisms and tone, but also not use language they wouldn't use. If they are not known to know about the concept in the question, don't offer an answer. Your answer should be no longer than 1 paragraph. And definitely try not to sound like anyone else. Don't repeat any slang or phrases already used. And if it is a question the original ${mentor} wouldn't have know the answer to, just say that you don't know, in the style of ${mentor}. And think about the time the person lived. Don't use terminology that they wouldn't have used.`
-
-  if (isLast) {
-    SystemPrompt += ` End your answer with something like I hope our answers help you out`;
-  } else {
-    SystemPrompt += ` Remember, this is a conversation, so you don't need a conclusion, but end your answer with a question related to the first question: "${question}".`;
-  }
-  return SystemPrompt;
-}
-
-async function main() {
-  const mentors = getMentors();
-  const question = process.argv[2];
-  let theConversation = `Here is the conversation so far.\nYou: ${question}\n`
-
-  for await (const mentor of mentors) {
-    const SystemPrompt = await getSystemPrompt(mentor, mentor === mentors[mentorCount - 1], question);
-    ollama.setModel(getMentorFileName(mentor));
-    ollama.setSystemPrompt(SystemPrompt);
-    let output = '';
-    process.stdout.write(`\n${mentor}: `);
-    for await (const chunk of ollama.streamingGenerate(theConversation + `Continue the conversation as if you were ${mentor} on the question "${question}".`)) {
-      if (chunk.response) {
-        output += chunk.response;
-        process.stdout.write(chunk.response);
-      } else {
-        process.stdout.write('\n');
-      }
-    }
-    theConversation += `${mentor}: ${output}\n\n`
-  }
-}
-
-main();
--- a/examples/typescript-mentors/package.json
+++ b/examples/typescript-mentors/package.json
@@ -1,7 +0,0 @@
-{
-  "dependencies": {
-    "fs": "^0.0.1-security",
-    "ollama-node": "^0.0.3",
-    "path": "^0.12.7"
-  }
-}
--- a/format/bytes.go
+++ b/format/bytes.go
@@ -1,16 +0,0 @@
-package format
-
-import "fmt"
-
-func HumanBytes(b int64) string {
-	switch {
-	case b > 1000*1000*1000:
-		return fmt.Sprintf("%d GB", b/1000/1000/1000)
-	case b > 1000*1000:
-		return fmt.Sprintf("%d MB", b/1000/1000)
-	case b > 1000:
-		return fmt.Sprintf("%d KB", b/1000)
-	default:
-		return fmt.Sprintf("%d B", b)
-	}
-}
--- a/format/time.go
+++ b/format/time.go
@@ -7,14 +7,26 @@ import (
 	"time"
 )

-// humanDuration returns a human-readable approximation of a
-// duration (eg. "About a minute", "4 hours ago", etc.).
-func humanDuration(d time.Duration) string {
+// HumanDuration returns a human-readable approximation of a duration
+// (eg. "About a minute", "4 hours ago", etc.).
+// Modified version of github.com/docker/go-units.HumanDuration
+func HumanDuration(d time.Duration) string {
+	return HumanDurationWithCase(d, true)
+}
+
+// HumanDurationWithCase returns a human-readable approximation of a
+// duration (eg. "About a minute", "4 hours ago", etc.). but allows
+// you to specify whether the first word should be capitalized
+// (eg. "About" vs. "about")
+func HumanDurationWithCase(d time.Duration, useCaps bool) string {
 	seconds := int(d.Seconds())

 	switch {
 	case seconds < 1:
-		return "Less than a second"
+		if useCaps {
+			return "Less than a second"
+		}
+		return "less than a second"
 	case seconds == 1:
 		return "1 second"
 	case seconds < 60:
@@ -24,7 +36,10 @@ func humanDuration(d time.Duration) string {
 	minutes := int(d.Minutes())
 	switch {
 	case minutes == 1:
-		return "About a minute"
+		if useCaps {
+			return "About a minute"
+		}
+		return "about a minute"
 	case minutes < 60:
 		return fmt.Sprintf("%d minutes", minutes)
 	}
@@ -32,7 +47,10 @@ func humanDuration(d time.Duration) string {
 	hours := int(math.Round(d.Hours()))
 	switch {
 	case hours == 1:
-		return "About an hour"
+		if useCaps {
+			return "About an hour"
+		}
+		return "about an hour"
 	case hours < 48:
 		return fmt.Sprintf("%d hours", hours)
 	case hours < 24*7*2:
@@ -47,22 +65,77 @@ func humanDuration(d time.Duration) string {
 }

 func HumanTime(t time.Time, zeroValue string) string {
-	return humanTime(t, zeroValue)
+	return humanTimeWithCase(t, zeroValue, true)
 }

 func HumanTimeLower(t time.Time, zeroValue string) string {
-	return strings.ToLower(humanTime(t, zeroValue))
+	return humanTimeWithCase(t, zeroValue, false)
 }

-func humanTime(t time.Time, zeroValue string) string {
+func humanTimeWithCase(t time.Time, zeroValue string, useCaps bool) string {
 	if t.IsZero() {
 		return zeroValue
 	}

 	delta := time.Since(t)
 	if delta < 0 {
-		return humanDuration(-delta) + " from now"
+		return HumanDurationWithCase(-delta, useCaps) + " from now"
+	}
+	return HumanDurationWithCase(delta, useCaps) + " ago"
+}
+
+// ExcatDuration returns a human readable hours/minutes/seconds or milliseconds format of a duration
+// the most precise level of duration is milliseconds
+func ExactDuration(d time.Duration) string {
+	if d.Seconds() < 1 {
+		if d.Milliseconds() == 1 {
+			return fmt.Sprintf("%d millisecond", d.Milliseconds())
+		}
+		return fmt.Sprintf("%d milliseconds", d.Milliseconds())
 	}

-	return humanDuration(delta) + " ago"
+	var readableDur strings.Builder
+
+	dur := d.String()
+
+	// split the default duration string format of 0h0m0s into something nicer to read
+	h := strings.Split(dur, "h")
+	if len(h) > 1 {
+		hours := h[0]
+		if hours == "1" {
+			readableDur.WriteString(fmt.Sprintf("%s hour ", hours))
+		} else {
+			readableDur.WriteString(fmt.Sprintf("%s hours ", hours))
+		}
+		dur = h[1]
+	}
+
+	m := strings.Split(dur, "m")
+	if len(m) > 1 {
+		mins := m[0]
+		switch mins {
+		case "0":
+			// skip
+		case "1":
+			readableDur.WriteString(fmt.Sprintf("%s minute ", mins))
+		default:
+			readableDur.WriteString(fmt.Sprintf("%s minutes ", mins))
+		}
+		dur = m[1]
+	}
+
+	s := strings.Split(dur, "s")
+	if len(s) > 0 {
+		sec := s[0]
+		switch sec {
+		case "0":
+			// skip
+		case "1":
+			readableDur.WriteString(fmt.Sprintf("%s second ", sec))
+		default:
+			readableDur.WriteString(fmt.Sprintf("%s seconds ", sec))
+		}
+	}
+
+	return strings.TrimSpace(readableDur.String())
 }
--- a/format/time_test.go
+++ b/format/time_test.go
@@ -11,25 +11,92 @@ func assertEqual(t *testing.T, a interface{}, b interface{}) {
 	}
 }

+func TestHumanDuration(t *testing.T) {
+	day := 24 * time.Hour
+	week := 7 * day
+	month := 30 * day
+	year := 365 * day
+
+	assertEqual(t, "Less than a second", HumanDuration(450*time.Millisecond))
+	assertEqual(t, "Less than a second", HumanDurationWithCase(450*time.Millisecond, true))
+	assertEqual(t, "less than a second", HumanDurationWithCase(450*time.Millisecond, false))
+	assertEqual(t, "1 second", HumanDuration(1*time.Second))
+	assertEqual(t, "45 seconds", HumanDuration(45*time.Second))
+	assertEqual(t, "46 seconds", HumanDuration(46*time.Second))
+	assertEqual(t, "59 seconds", HumanDuration(59*time.Second))
+	assertEqual(t, "About a minute", HumanDuration(60*time.Second))
+	assertEqual(t, "About a minute", HumanDurationWithCase(1*time.Minute, true))
+	assertEqual(t, "about a minute", HumanDurationWithCase(1*time.Minute, false))
+	assertEqual(t, "3 minutes", HumanDuration(3*time.Minute))
+	assertEqual(t, "35 minutes", HumanDuration(35*time.Minute))
+	assertEqual(t, "35 minutes", HumanDuration(35*time.Minute+40*time.Second))
+	assertEqual(t, "45 minutes", HumanDuration(45*time.Minute))
+	assertEqual(t, "45 minutes", HumanDuration(45*time.Minute+40*time.Second))
+	assertEqual(t, "46 minutes", HumanDuration(46*time.Minute))
+	assertEqual(t, "59 minutes", HumanDuration(59*time.Minute))
+	assertEqual(t, "About an hour", HumanDuration(1*time.Hour))
+	assertEqual(t, "About an hour", HumanDurationWithCase(1*time.Hour+29*time.Minute, true))
+	assertEqual(t, "about an hour", HumanDurationWithCase(1*time.Hour+29*time.Minute, false))
+	assertEqual(t, "2 hours", HumanDuration(1*time.Hour+31*time.Minute))
+	assertEqual(t, "2 hours", HumanDuration(1*time.Hour+59*time.Minute))
+	assertEqual(t, "3 hours", HumanDuration(3*time.Hour))
+	assertEqual(t, "3 hours", HumanDuration(3*time.Hour+29*time.Minute))
+	assertEqual(t, "4 hours", HumanDuration(3*time.Hour+31*time.Minute))
+	assertEqual(t, "4 hours", HumanDuration(3*time.Hour+59*time.Minute))
+	assertEqual(t, "4 hours", HumanDuration(3*time.Hour+60*time.Minute))
+	assertEqual(t, "24 hours", HumanDuration(24*time.Hour))
+	assertEqual(t, "36 hours", HumanDuration(1*day+12*time.Hour))
+	assertEqual(t, "2 days", HumanDuration(2*day))
+	assertEqual(t, "7 days", HumanDuration(7*day))
+	assertEqual(t, "13 days", HumanDuration(13*day+5*time.Hour))
+	assertEqual(t, "2 weeks", HumanDuration(2*week))
+	assertEqual(t, "2 weeks", HumanDuration(2*week+4*day))
+	assertEqual(t, "3 weeks", HumanDuration(3*week))
+	assertEqual(t, "4 weeks", HumanDuration(4*week))
+	assertEqual(t, "4 weeks", HumanDuration(4*week+3*day))
+	assertEqual(t, "4 weeks", HumanDuration(1*month))
+	assertEqual(t, "6 weeks", HumanDuration(1*month+2*week))
+	assertEqual(t, "2 months", HumanDuration(2*month))
+	assertEqual(t, "2 months", HumanDuration(2*month+2*week))
+	assertEqual(t, "3 months", HumanDuration(3*month))
+	assertEqual(t, "3 months", HumanDuration(3*month+1*week))
+	assertEqual(t, "5 months", HumanDuration(5*month+2*week))
+	assertEqual(t, "13 months", HumanDuration(13*month))
+	assertEqual(t, "23 months", HumanDuration(23*month))
+	assertEqual(t, "24 months", HumanDuration(24*month))
+	assertEqual(t, "2 years", HumanDuration(24*month+2*week))
+	assertEqual(t, "3 years", HumanDuration(3*year+2*month))
+}
+
 func TestHumanTime(t *testing.T) {
 	now := time.Now()

 	t.Run("zero value", func(t *testing.T) {
 		assertEqual(t, HumanTime(time.Time{}, "never"), "never")
 	})
-
 	t.Run("time in the future", func(t *testing.T) {
 		v := now.Add(48 * time.Hour)
 		assertEqual(t, HumanTime(v, ""), "2 days from now")
 	})
-
 	t.Run("time in the past", func(t *testing.T) {
 		v := now.Add(-48 * time.Hour)
 		assertEqual(t, HumanTime(v, ""), "2 days ago")
 	})
-
-	t.Run("soon", func(t *testing.T) {
-		v := now.Add(800*time.Millisecond)
-		assertEqual(t, HumanTime(v, ""), "Less than a second from now")
-	})
+}
+
+func TestExactDuration(t *testing.T) {
+	assertEqual(t, "1 millisecond", ExactDuration(1*time.Millisecond))
+	assertEqual(t, "10 milliseconds", ExactDuration(10*time.Millisecond))
+	assertEqual(t, "1 second", ExactDuration(1*time.Second))
+	assertEqual(t, "10 seconds", ExactDuration(10*time.Second))
+	assertEqual(t, "1 minute", ExactDuration(1*time.Minute))
+	assertEqual(t, "10 minutes", ExactDuration(10*time.Minute))
+	assertEqual(t, "1 hour", ExactDuration(1*time.Hour))
+	assertEqual(t, "10 hours", ExactDuration(10*time.Hour))
+	assertEqual(t, "1 hour 1 second", ExactDuration(1*time.Hour+1*time.Second))
+	assertEqual(t, "1 hour 10 seconds", ExactDuration(1*time.Hour+10*time.Second))
+	assertEqual(t, "1 hour 1 minute", ExactDuration(1*time.Hour+1*time.Minute))
+	assertEqual(t, "1 hour 10 minutes", ExactDuration(1*time.Hour+10*time.Minute))
+	assertEqual(t, "1 hour 1 minute 1 second", ExactDuration(1*time.Hour+1*time.Minute+1*time.Second))
+	assertEqual(t, "10 hours 10 minutes 10 seconds", ExactDuration(10*time.Hour+10*time.Minute+10*time.Second))
 }
--- a/go.mod
+++ b/go.mod
@@ -10,7 +10,6 @@ require (
 	github.com/olekukonko/tablewriter v0.0.5
 	github.com/pdevine/readline v1.5.2
 	github.com/spf13/cobra v1.7.0
-	golang.org/x/sync v0.3.0
 )

 require github.com/rivo/uniseg v0.2.0 // indirect
--- a/go.sum
+++ b/go.sum
@@ -125,8 +125,6 @@ golang.org/x/exp v0.0.0-20230817173708-d852ddb80c63/go.mod h1:0v4NqG35kSWCMzLaMe
 golang.org/x/net v0.0.0-20210226172049-e18ecbb05110/go.mod h1:m0MpNAwzfU5UDzcl9v0D8zg8gWTRqZa9RBIspLL5mdg=
 golang.org/x/net v0.10.0 h1:X2//UzNDwYmtCLn7To6G58Wr6f5ahEAQgKNzv9Y951M=
 golang.org/x/net v0.10.0/go.mod h1:0qNGK6F8kojg2nk9dLZ2mShWaEBan6FAoqfSigmmuDg=
-golang.org/x/sync v0.3.0 h1:ftCYgMx6zT/asHUrPw8BLLscYtGznsLAnjq5RH9P66E=
-golang.org/x/sync v0.3.0/go.mod h1:FU7BRWz2tNW+3quACPkgCx/L+uEAv1htQ0V83Z9Rj+Y=
 golang.org/x/sys v0.0.0-20201119102817-f84b799fce68/go.mod h1:h1NjWce9XRLGQEsW7wpKNCjG9DtNlClVuFLEZdDNbEs=
 golang.org/x/sys v0.0.0-20210615035016-665e8c7367d1/go.mod h1:oPkhp1MJrh7nUepCBck5+mAzfO9JrbApNNgaTdGDITg=
 golang.org/x/sys v0.0.0-20210630005230-0f9fa26af87c/go.mod h1:oPkhp1MJrh7nUepCBck5+mAzfO9JrbApNNgaTdGDITg=
--- a/llm/falcon.go
+++ b/llm/falcon.go
@@ -1,5 +1,7 @@
 package llm

+const ModelFamilyFalcon = "falcon"
+
 const (
 	falconModelType7B   = 32
 	falconModelType40B  = 60
@@ -15,6 +17,6 @@ func falconModelType(numLayer uint32) string {
 	case 80:
 		return "180B"
 	default:
-		return "unknown"
+		return "Unknown"
 	}
 }
--- a/llm/ggml.go
+++ b/llm/ggml.go
@@ -69,7 +69,7 @@ func fileType(fileType uint32) string {
 	case fileTypeQ6_K:
 		return "Q6_K"
 	default:
-		return "unknown"
+		return "Unknown"
 	}
 }

--- a/llm/gguf.go
+++ b/llm/gguf.go
@@ -109,13 +109,9 @@ func (llm *ggufModel) ModelType() string {
 		if blocks, ok := llm.kv["falcon.block_count"].(uint32); ok {
 			return falconModelType(blocks)
 		}
-	case "starcoder":
-		if blocks, ok := llm.kv["starcoder.block_count"].(uint32); ok {
-			return starCoderModelType(blocks)
-		}
 	}

-	return "unknown"
+	return "Unknown"
 }

 func (llm *ggufModel) FileType() string {
@@ -124,7 +120,7 @@ func (llm *ggufModel) FileType() string {
 		return fileType(t)
 	}

-	return "unknown"
+	return "Unknown"
 }

 func (llm *ggufModel) Decode(r io.Reader) error {
--- a/llm/llama.cpp/generate_darwin_amd64.go
+++ b/llm/llama.cpp/generate_darwin_amd64.go
@@ -9,10 +9,8 @@ package llm
 //go:generate git -C ggml apply ../patches/0004-metal-add-missing-barriers-for-mul-mat-2699.patch
 //go:generate cmake -S ggml -B ggml/build/cpu -DLLAMA_ACCELERATE=on -DLLAMA_K_QUANTS=on -DCMAKE_SYSTEM_PROCESSOR=x86_64 -DCMAKE_OSX_ARCHITECTURES=x86_64 -DCMAKE_OSX_DEPLOYMENT_TARGET=11.0
 //go:generate cmake --build ggml/build/cpu --target server --config Release
-//go:generate mv ggml/build/cpu/bin/server ggml/build/cpu/bin/ollama-runner

 //go:generate git submodule update --force gguf
 //go:generate git -C gguf apply ../patches/0001-remove-warm-up-logging.patch
 //go:generate cmake -S gguf -B gguf/build/cpu -DLLAMA_ACCELERATE=on -DLLAMA_K_QUANTS=on -DCMAKE_SYSTEM_PROCESSOR=x86_64 -DCMAKE_OSX_ARCHITECTURES=x86_64 -DCMAKE_OSX_DEPLOYMENT_TARGET=11.0
 //go:generate cmake --build gguf/build/cpu --target server --config Release
-//go:generate mv gguf/build/cpu/bin/server gguf/build/cpu/bin/ollama-runner
--- a/llm/llama.cpp/generate_darwin_arm64.go
+++ b/llm/llama.cpp/generate_darwin_arm64.go
@@ -9,10 +9,8 @@ package llm
 //go:generate git -C ggml apply ../patches/0004-metal-add-missing-barriers-for-mul-mat-2699.patch
 //go:generate cmake -S ggml -B ggml/build/metal -DLLAMA_METAL=on -DLLAMA_ACCELERATE=on -DLLAMA_K_QUANTS=on -DCMAKE_SYSTEM_PROCESSOR=arm64 -DCMAKE_OSX_ARCHITECTURES=arm64 -DCMAKE_OSX_DEPLOYMENT_TARGET=11.0
 //go:generate cmake --build ggml/build/metal --target server --config Release
-//go:generate mv ggml/build/metal/bin/server ggml/build/metal/bin/ollama-runner

 //go:generate git submodule update --force gguf
 //go:generate git -C gguf apply ../patches/0001-remove-warm-up-logging.patch
 //go:generate cmake -S gguf -B gguf/build/metal -DLLAMA_METAL=on -DLLAMA_ACCELERATE=on -DLLAMA_K_QUANTS=on -DCMAKE_SYSTEM_PROCESSOR=arm64 -DCMAKE_OSX_ARCHITECTURES=arm64 -DCMAKE_OSX_DEPLOYMENT_TARGET=11.0
 //go:generate cmake --build gguf/build/metal --target server --config Release
-//go:generate mv gguf/build/metal/bin/server gguf/build/metal/bin/ollama-runner
--- a/llm/llama.cpp/generate_linux.go
+++ b/llm/llama.cpp/generate_linux.go
@@ -9,18 +9,14 @@ package llm
 //go:generate git -C ggml apply ../patches/0001-copy-cuda-runtime-libraries.patch
 //go:generate cmake -S ggml -B ggml/build/cpu -DLLAMA_K_QUANTS=on
 //go:generate cmake --build ggml/build/cpu --target server --config Release
-//go:generate mv ggml/build/cpu/bin/server ggml/build/cpu/bin/ollama-runner

 //go:generate git submodule update --force gguf
 //go:generate git -C gguf apply ../patches/0001-copy-cuda-runtime-libraries.patch
 //go:generate git -C gguf apply ../patches/0001-remove-warm-up-logging.patch
 //go:generate cmake -S gguf -B gguf/build/cpu -DLLAMA_K_QUANTS=on
 //go:generate cmake --build gguf/build/cpu --target server --config Release
-//go:generate mv gguf/build/cpu/bin/server gguf/build/cpu/bin/ollama-runner

 //go:generate cmake -S ggml -B ggml/build/cuda -DLLAMA_CUBLAS=on -DLLAMA_ACCELERATE=on -DLLAMA_K_QUANTS=on
 //go:generate cmake --build ggml/build/cuda --target server --config Release
-//go:generate mv ggml/build/cuda/bin/server ggml/build/cuda/bin/ollama-runner
 //go:generate cmake -S gguf -B gguf/build/cuda -DLLAMA_CUBLAS=on -DLLAMA_ACCELERATE=on -DLLAMA_K_QUANTS=on
 //go:generate cmake --build gguf/build/cuda --target server --config Release
-//go:generate mv gguf/build/cuda/bin/server gguf/build/cuda/bin/ollama-runner
--- a/llm/llama.cpp/generate_windows.go
+++ b/llm/llama.cpp/generate_windows.go
@@ -7,10 +7,8 @@ package llm
 //go:generate git -C ggml apply ../patches/0002-34B-model-support.patch
 //go:generate cmake -S ggml -B ggml/build/cpu -DLLAMA_K_QUANTS=on
 //go:generate cmake --build ggml/build/cpu --target server --config Release
-//go:generate cmd /c move ggml\build\cpu\bin\Release\server.exe ggml\build\cpu\bin\Release\ollama-runner.exe

 //go:generate git submodule update --force gguf
 //go:generate git -C gguf apply ../patches/0001-remove-warm-up-logging.patch
 //go:generate cmake -S gguf -B gguf/build/cpu -DLLAMA_K_QUANTS=on
 //go:generate cmake --build gguf/build/cpu --target server --config Release
-//go:generate cmd /c move gguf\build\cpu\bin\Release\server.exe gguf\build\cpu\bin\Release\ollama-runner.exe
--- a/llm/llama.go
+++ b/llm/llama.go
@@ -20,7 +20,6 @@ import (
 	"runtime"
 	"strconv"
 	"strings"
-	"sync"
 	"time"

 	"github.com/jmorganca/ollama/api"
@@ -42,23 +41,23 @@ func chooseRunners(workDir, runnerType string) []ModelRunner {
 	switch runtime.GOOS {
 	case "darwin":
 		runners = []string{
-			path.Join(buildPath, "metal", "bin", "ollama-runner"),
-			path.Join(buildPath, "cpu", "bin", "ollama-runner"),
+			path.Join(buildPath, "metal", "bin", "server"),
+			path.Join(buildPath, "cpu", "bin", "server"),
 		}
 	case "linux":
 		runners = []string{
-			path.Join(buildPath, "cuda", "bin", "ollama-runner"),
-			path.Join(buildPath, "cpu", "bin", "ollama-runner"),
+			path.Join(buildPath, "cuda", "bin", "server"),
+			path.Join(buildPath, "cpu", "bin", "server"),
 		}
 	case "windows":
 		// TODO: select windows GPU runner here when available
 		runners = []string{
-			path.Join(buildPath, "cpu", "bin", "Release", "ollama-runner.exe"),
+			path.Join(buildPath, "cpu", "bin", "Release", "server.exe"),
 		}
 	default:
 		log.Printf("unknown OS, running on CPU: %s", runtime.GOOS)
 		runners = []string{
-			path.Join(buildPath, "cpu", "bin", "ollama-runner"),
+			path.Join(buildPath, "cpu", "bin", "server"),
 		}
 	}

@@ -144,7 +143,7 @@ func llamaModelType(numLayer uint32) string {
 	case 80:
 		return "65B"
 	default:
-		return "unknown"
+		return "Unknown"
 	}
 }

@@ -178,12 +177,9 @@ type llamaHyperparameters struct {
 }

 type Running struct {
-	Port     int
-	Cmd      *exec.Cmd
-	Cancel   context.CancelFunc
-	exitOnce sync.Once
-	exitCh   chan error // channel to receive the exit status of the subprocess
-	exitErr  error      // error returned by the subprocess
+	Port   int
+	Cmd    *exec.Cmd
+	Cancel context.CancelFunc
 }

 type llama struct {
@@ -195,7 +191,7 @@ var errNoGPU = errors.New("nvidia-smi command failed")

 // CheckVRAM returns the available VRAM in MiB on Linux machines with NVIDIA GPUs
 func CheckVRAM() (int64, error) {
-	cmd := exec.Command("nvidia-smi", "--query-gpu=memory.free", "--format=csv,noheader,nounits")
+	cmd := exec.Command("nvidia-smi", "--query-gpu=memory.total", "--format=csv,noheader,nounits")
 	var stdout bytes.Buffer
 	cmd.Stdout = &stdout
 	err := cmd.Run()
@@ -203,7 +199,7 @@ func CheckVRAM() (int64, error) {
 		return 0, errNoGPU
 	}

-	var free int64
+	var total int64
 	scanner := bufio.NewScanner(&stdout)
 	for scanner.Scan() {
 		line := scanner.Text()
@@ -212,16 +208,17 @@ func CheckVRAM() (int64, error) {
 			return 0, fmt.Errorf("failed to parse available VRAM: %v", err)
 		}

-		free += vram
+		total += vram
 	}

-	return free, nil
+	return total, nil
 }

 func NumGPU(numLayer, fileSizeBytes int64, opts api.Options) int {
 	if opts.NumGPU != -1 {
 		return opts.NumGPU
 	}
+	n := 1 // default to enable metal on macOS
 	if runtime.GOOS == "linux" {
 		vramMib, err := CheckVRAM()
 		if err != nil {
@@ -232,41 +229,21 @@ func NumGPU(numLayer, fileSizeBytes int64, opts api.Options) int {
 			return 0
 		}

-		freeVramBytes := int64(vramMib) * 1024 * 1024 // 1 MiB = 1024^2 bytes
+		totalVramBytes := int64(vramMib) * 1024 * 1024 // 1 MiB = 1024^2 bytes

 		// Calculate bytes per layer
 		// TODO: this is a rough heuristic, better would be to calculate this based on number of layers and context size
 		bytesPerLayer := fileSizeBytes / numLayer

-		// max number of layers we can fit in VRAM, subtract 5% to prevent consuming all available VRAM and running out of memory
-		layers := int(freeVramBytes/bytesPerLayer) * 95 / 100
-		log.Printf("%d MiB VRAM available, loading up to %d GPU layers", vramMib, layers)
+		// set n to the max number of layers we can fit in VRAM
+		return int(totalVramBytes / bytesPerLayer)

-		return layers
+		log.Printf("%d MiB VRAM available, loading up to %d GPU layers", vramMib, n)
 	}
 	// default to enable metal on macOS
 	return 1
 }

-// StatusWriter is a writer that captures error messages from the llama runner process
-type StatusWriter struct {
-	ErrCh chan error
-}
-
-func NewStatusWriter() *StatusWriter {
-	return &StatusWriter{
-		ErrCh: make(chan error, 1),
-	}
-}
-
-func (w *StatusWriter) Write(b []byte) (int, error) {
-	if _, after, ok := bytes.Cut(b, []byte("error:")); ok {
-		err := fmt.Errorf("llama runner: %s", after)
-		w.ErrCh <- err
-	}
-	return os.Stderr.Write(b)
-}
-
 func newLlama(model string, adapters []string, runners []ModelRunner, numLayers int64, opts api.Options) (*llama, error) {
 	fileInfo, err := os.Stat(model)
 	if err != nil {
@@ -313,8 +290,6 @@ func newLlama(model string, adapters []string, runners []ModelRunner, numLayers
 		params = append(params, "--numa")
 	}

-	var runnerErr error
-
 	// start the llama.cpp server with a retry in case the port is already in use
 	for _, runner := range runners {
 		if _, err := os.Stat(runner.Path); err != nil {
@@ -331,10 +306,9 @@ func newLlama(model string, adapters []string, runners []ModelRunner, numLayers
 		)
 		cmd.Env = append(os.Environ(), fmt.Sprintf("LD_LIBRARY_PATH=%s", filepath.Dir(runner.Path)))
 		cmd.Stdout = os.Stderr
-		statusWriter := NewStatusWriter()
-		cmd.Stderr = statusWriter
+		cmd.Stderr = os.Stderr

-		llm := &llama{Options: opts, Running: Running{Port: port, Cmd: cmd, Cancel: cancel, exitCh: make(chan error)}}
+		llm := &llama{Options: opts, Running: Running{Port: port, Cmd: cmd, Cancel: cancel}}

 		log.Print("starting llama runner")
 		if err := llm.Cmd.Start(); err != nil {
@@ -342,30 +316,19 @@ func newLlama(model string, adapters []string, runners []ModelRunner, numLayers
 			continue
 		}

-		// monitor the llama runner process and signal when it exits
+		// monitor the command, it is blocking, so if it exits we need to capture that
 		go func() {
-			err := llm.Cmd.Wait()
-			llm.exitErr = err
-			// llm.Cmd.Wait() can only be called once, use this exit channel to signal that the process has exited
-			llm.exitOnce.Do(func() {
-				close(llm.exitCh)
-			})
+			err := llm.Cmd.Wait() // this will block until the command exits
+			if err != nil {
+				log.Printf("llama runner exited with error: %v", err)
+			} else {
+				log.Printf("llama runner exited")
+			}
 		}()

 		if err := waitForServer(llm); err != nil {
 			log.Printf("error starting llama runner: %v", err)
 			llm.Close()
-
-			// default the runnerErr to the error returned by the most recent llama runner process
-			runnerErr = err
-
-			// capture the error directly from the runner process, if any
-			select {
-			case runnerErr = <-statusWriter.ErrCh:
-			default:
-				// the runner process probably timed out
-			}
-
 			// try again
 			continue
 		}
@@ -374,54 +337,37 @@ func newLlama(model string, adapters []string, runners []ModelRunner, numLayers
 		return llm, nil
 	}

-	if runnerErr != nil {
-		// this is the error returned from the llama runner process that failed most recently
-		return nil, runnerErr
-	}
-
 	return nil, fmt.Errorf("failed to start a llama runner")
 }

 func waitForServer(llm *llama) error {
+	// wait for the server to start responding
 	start := time.Now()
-	expiresAt := time.Now().Add(3 * time.Minute) // be generous with timeout, large models can take a while to load
+	expiresAt := time.Now().Add(2 * time.Minute) // be generous with timeout, large models can take a while to load
 	ticker := time.NewTicker(200 * time.Millisecond)
-	defer ticker.Stop()

 	log.Print("waiting for llama runner to start responding")
-	for {
-		select {
-		case <-llm.exitCh:
-			// failed to start subprocess
-			return fmt.Errorf("llama runner process has terminated")
-		case <-ticker.C:
-			if time.Now().After(expiresAt) {
-				// timeout
-				return fmt.Errorf("timed out waiting for llama runner to start")
-			}
+	for range ticker.C {
+		if time.Now().After(expiresAt) {
+			return fmt.Errorf("llama runner did not start within alloted time, retrying")
+		}

-			if err := llm.Ping(context.Background()); err == nil {
-				// success
-				log.Printf("llama runner started in %f seconds", time.Since(start).Seconds())
-				return nil
-			}
+		// check if the server process has terminated
+		if llm.Cmd.ProcessState != nil && llm.Cmd.ProcessState.Exited() {
+			return fmt.Errorf("llama runner process has terminated")
+		}
+
+		if err := llm.Ping(context.Background()); err == nil {
+			break
 		}
 	}
+
+	log.Printf("llama runner started in %f seconds", time.Since(start).Seconds())
+	return nil
 }

 func (llm *llama) Close() {
-	// signal the sub-process to terminate
 	llm.Cancel()
-
-	// wait for the command to exit to prevent race conditions with the next run
-	<-llm.exitCh
-	err := llm.exitErr
-
-	if err != nil {
-		log.Printf("llama runner stopped with error: %v", err)
-	} else {
-		log.Print("llama runner stopped successfully")
-	}
 }

 func (llm *llama) SetOptions(opts api.Options) {
@@ -471,29 +417,30 @@ type Prediction struct {
 }

 type PredictRequest struct {
-	Prompt           string   `json:"prompt"`
-	Stream           bool     `json:"stream"`
-	NPredict         int      `json:"n_predict"`
-	NKeep            int      `json:"n_keep"`
-	Temperature      float32  `json:"temperature"`
-	TopK             int      `json:"top_k"`
-	TopP             float32  `json:"top_p"`
-	TfsZ             float32  `json:"tfs_z"`
-	TypicalP         float32  `json:"typical_p"`
-	RepeatLastN      int      `json:"repeat_last_n"`
-	RepeatPenalty    float32  `json:"repeat_penalty"`
-	PresencePenalty  float32  `json:"presence_penalty"`
-	FrequencyPenalty float32  `json:"frequency_penalty"`
-	Mirostat         int      `json:"mirostat"`
-	MirostatTau      float32  `json:"mirostat_tau"`
-	MirostatEta      float32  `json:"mirostat_eta"`
-	PenalizeNl       bool     `json:"penalize_nl"`
-	Seed             int      `json:"seed"`
-	Stop             []string `json:"stop,omitempty"`
+	Stream           bool            `json:"stream"`
+	NPredict         int             `json:"n_predict,omitempty"`
+	TopK             int             `json:"top_k,omitempty"`
+	TopP             float32         `json:"top_p,omitempty"`
+	TfsZ             float32         `json:"tfs_z,omitempty"`
+	TypicalP         float32         `json:"typical_p,omitempty"`
+	RepeatLastN      int             `json:"repeat_last_n,omitempty"`
+	Temperature      float32         `json:"temperature,omitempty"`
+	RepeatPenalty    float32         `json:"repeat_penalty,omitempty"`
+	PresencePenalty  float32         `json:"presence_penalty,omitempty"`
+	FrequencyPenalty float32         `json:"frequency_penalty,omitempty"`
+	Mirostat         int             `json:"mirostat,omitempty"`
+	MirostatTau      float32         `json:"mirostat_tau,omitempty"`
+	MirostatEta      float32         `json:"mirostat_eta,omitempty"`
+	PenalizeNl       bool            `json:"penalize_nl,omitempty"`
+	NKeep            int             `json:"n_keep,omitempty"`
+	Seed             int             `json:"seed,omitempty"`
+	Prompt           string          `json:"prompt,omitempty"`
+	NProbs           int             `json:"n_probs,omitempty"`
+	LogitBias        map[int]float32 `json:"logit_bias,omitempty"`
+	IgnoreEos        bool            `json:"ignore_eos,omitempty"`
+	Stop             []string        `json:"stop,omitempty"`
 }

-const maxBufferSize = 512 * 1000 // 512KB
-
 func (llm *llama) Predict(ctx context.Context, prevContext []int, prompt string, fn func(api.GenerateResponse)) error {
 	prevConvo, err := llm.Decode(ctx, prevContext)
 	if err != nil {
@@ -523,10 +470,8 @@ func (llm *llama) Predict(ctx context.Context, prevContext []int, prompt string,
 		MirostatTau:      llm.MirostatTau,
 		MirostatEta:      llm.MirostatEta,
 		PenalizeNl:       llm.PenalizeNewline,
-		Seed:             llm.Seed,
 		Stop:             llm.Stop,
 	}
-
 	data, err := json.Marshal(predReq)
 	if err != nil {
 		return fmt.Errorf("error marshaling data: %v", err)
@@ -554,9 +499,6 @@ func (llm *llama) Predict(ctx context.Context, prevContext []int, prompt string,
 	}

 	scanner := bufio.NewScanner(resp.Body)
-	// increase the buffer size to avoid running out of space
-	buf := make([]byte, 0, maxBufferSize)
-	scanner.Buffer(buf, maxBufferSize)
 	for scanner.Scan() {
 		select {
 		case <-ctx.Done():
--- a/llm/llm.go
+++ b/llm/llm.go
@@ -5,7 +5,6 @@ import (
 	"fmt"
 	"log"
 	"os"
-	"runtime"

 	"github.com/pbnjay/memory"

@@ -38,55 +37,53 @@ func New(workDir, model string, adapters []string, opts api.Options) (LLM, error
 		return nil, err
 	}

-	if runtime.GOOS == "darwin" {
-		switch ggml.FileType() {
-		case "Q8_0":
-			if ggml.Name() != "gguf" && opts.NumGPU != 0 {
-				// GGML Q8_0 do not support Metal API and will
-				// cause the runner to segmentation fault so disable GPU
-				log.Printf("WARNING: GPU disabled for F32, Q5_0, Q5_1, and Q8_0")
-				opts.NumGPU = 0
-			}
-		case "F32", "Q5_0", "Q5_1":
-			if opts.NumGPU != 0 {
-				// F32, Q5_0, Q5_1, and Q8_0 do not support Metal API and will
-				// cause the runner to segmentation fault so disable GPU
-				log.Printf("WARNING: GPU disabled for F32, Q5_0, Q5_1, and Q8_0")
-				opts.NumGPU = 0
-			}
+	switch ggml.FileType() {
+	case "Q8_0":
+		if ggml.Name() != "gguf" && opts.NumGPU != 0 {
+			// GGML Q8_0 do not support Metal API and will
+			// cause the runner to segmentation fault so disable GPU
+			log.Printf("WARNING: GPU disabled for F32, Q5_0, Q5_1, and Q8_0")
+			opts.NumGPU = 0
+		}
+	case "F32", "Q5_0", "Q5_1":
+		if opts.NumGPU != 0 {
+			// F32, Q5_0, Q5_1, and Q8_0 do not support Metal API and will
+			// cause the runner to segmentation fault so disable GPU
+			log.Printf("WARNING: GPU disabled for F32, Q5_0, Q5_1, and Q8_0")
+			opts.NumGPU = 0
 		}
 	}

 	totalResidentMemory := memory.TotalMemory()
 	switch ggml.ModelType() {
 	case "3B", "7B":
-		if ggml.FileType() == "F16" && totalResidentMemory < 16*1000*1000 {
-			return nil, fmt.Errorf("F16 model requires at least 16 GB of memory")
-		} else if totalResidentMemory < 8*1000*1000 {
-			return nil, fmt.Errorf("model requires at least 8 GB of memory")
+		if ggml.FileType() == "F16" && totalResidentMemory < 16*1024*1024 {
+			return nil, fmt.Errorf("F16 model requires at least 16GB of memory")
+		} else if totalResidentMemory < 8*1024*1024 {
+			return nil, fmt.Errorf("model requires at least 8GB of memory")
 		}
 	case "13B":
-		if ggml.FileType() == "F16" && totalResidentMemory < 32*1000*1000 {
-			return nil, fmt.Errorf("F16 model requires at least 32 GB of memory")
-		} else if totalResidentMemory < 16*1000*1000 {
-			return nil, fmt.Errorf("model requires at least 16 GB of memory")
+		if ggml.FileType() == "F16" && totalResidentMemory < 32*1024*1024 {
+			return nil, fmt.Errorf("F16 model requires at least 32GB of memory")
+		} else if totalResidentMemory < 16*1024*1024 {
+			return nil, fmt.Errorf("model requires at least 16GB of memory")
 		}
 	case "30B", "34B", "40B":
-		if ggml.FileType() == "F16" && totalResidentMemory < 64*1000*1000 {
-			return nil, fmt.Errorf("F16 model requires at least 64 GB of memory")
-		} else if totalResidentMemory < 32*1000*1000 {
-			return nil, fmt.Errorf("model requires at least 32 GB of memory")
+		if ggml.FileType() == "F16" && totalResidentMemory < 64*1024*1024 {
+			return nil, fmt.Errorf("F16 model requires at least 64GB of memory")
+		} else if totalResidentMemory < 32*1024*1024 {
+			return nil, fmt.Errorf("model requires at least 32GB of memory")
 		}
 	case "65B", "70B":
-		if ggml.FileType() == "F16" && totalResidentMemory < 128*1000*1000 {
-			return nil, fmt.Errorf("F16 model requires at least 128 GB of memory")
-		} else if totalResidentMemory < 64*1000*1000 {
-			return nil, fmt.Errorf("model requires at least 64 GB of memory")
+		if ggml.FileType() == "F16" && totalResidentMemory < 128*1024*1024 {
+			return nil, fmt.Errorf("F16 model requires at least 128GB of memory")
+		} else if totalResidentMemory < 64*1024*1024 {
+			return nil, fmt.Errorf("model requires at least 64GB of memory")
 		}
 	case "180B":
-		if ggml.FileType() == "F16" && totalResidentMemory < 512*1000*1000 {
+		if ggml.FileType() == "F16" && totalResidentMemory < 512*1024*1024 {
 			return nil, fmt.Errorf("F16 model requires at least 512GB of memory")
-		} else if totalResidentMemory < 128*1000*1000 {
+		} else if totalResidentMemory < 128*1024*1024 {
 			return nil, fmt.Errorf("model requires at least 128GB of memory")
 		}
 	}
--- a/llm/starcoder.go
+++ b/llm/starcoder.go
@@ -1,23 +0,0 @@
-package llm
-
-const (
-	starCoderModelType1B  = 24
-	starCoderModelType3B  = 36
-	starCoderModelType7B  = 42
-	starCoderModelType15B = 40
-)
-
-func starCoderModelType(numLayer uint32) string {
-	switch numLayer {
-	case 24:
-		return "1B"
-	case 36:
-		return "3B"
-	case 42:
-		return "7B"
-	case 40:
-		return "15B"
-	default:
-		return "unknown"
-	}
-}
--- a/scripts/build_docker.sh
+++ b/scripts/build_docker.sh
--- a/server/download.go
+++ b/server/download.go
@@ -2,309 +2,38 @@ package server

 import (
 	"context"
-	"encoding/json"
 	"errors"
 	"fmt"
 	"io"
 	"log"
 	"net/http"
-	"net/url"
 	"os"
 	"path/filepath"
 	"strconv"
-	"strings"
 	"sync"
-	"sync/atomic"
 	"time"

-	"golang.org/x/sync/errgroup"
-
 	"github.com/jmorganca/ollama/api"
-	"github.com/jmorganca/ollama/format"
 )

-var blobDownloadManager sync.Map
-
-type blobDownload struct {
-	Name   string
-	Digest string
-
+type FileDownload struct {
+	Digest    string
+	FilePath  string
 	Total     int64
-	Completed atomic.Int64
-
-	Parts []*blobDownloadPart
-
-	context.CancelFunc
-
-	done       bool
-	err        error
-	references atomic.Int32
-}
-
-type blobDownloadPart struct {
-	N         int
-	Offset    int64
-	Size      int64
 	Completed int64
-
-	*blobDownload `json:"-"`
 }

-const (
-	numDownloadParts          = 64
-	minDownloadPartSize int64 = 32 * 1000 * 1000
-	maxDownloadPartSize int64 = 256 * 1000 * 1000
-)
-
-func (p *blobDownloadPart) Name() string {
-	return strings.Join([]string{
-		p.blobDownload.Name, "partial", strconv.Itoa(p.N),
-	}, "-")
-}
-
-func (p *blobDownloadPart) StartsAt() int64 {
-	return p.Offset + p.Completed
-}
-
-func (p *blobDownloadPart) StopsAt() int64 {
-	return p.Offset + p.Size
-}
-
-func (b *blobDownload) Prepare(ctx context.Context, requestURL *url.URL, opts *RegistryOptions) error {
-	partFilePaths, err := filepath.Glob(b.Name + "-partial-*")
-	if err != nil {
-		return err
-	}
-
-	for _, partFilePath := range partFilePaths {
-		part, err := b.readPart(partFilePath)
-		if err != nil {
-			return err
-		}
-
-		b.Total += part.Size
-		b.Completed.Add(part.Completed)
-		b.Parts = append(b.Parts, part)
-	}
-
-	if len(b.Parts) == 0 {
-		resp, err := makeRequest(ctx, "HEAD", requestURL, nil, nil, opts)
-		if err != nil {
-			return err
-		}
-		defer resp.Body.Close()
-
-		if resp.StatusCode >= http.StatusBadRequest {
-			body, _ := io.ReadAll(resp.Body)
-			return fmt.Errorf("registry responded with code %d: %v", resp.StatusCode, string(body))
-		}
-
-		b.Total, _ = strconv.ParseInt(resp.Header.Get("Content-Length"), 10, 64)
-
-		var size = b.Total / numDownloadParts
-		switch {
-		case size < minDownloadPartSize:
-			size = minDownloadPartSize
-		case size > maxDownloadPartSize:
-			size = maxDownloadPartSize
-		}
-
-		var offset int64
-		for offset < b.Total {
-			if offset+size > b.Total {
-				size = b.Total - offset
-			}
-
-			if err := b.newPart(offset, size); err != nil {
-				return err
-			}
-
-			offset += size
-		}
-	}
-
-	log.Printf("downloading %s in %d %s part(s)", b.Digest[7:19], len(b.Parts), format.HumanBytes(b.Parts[0].Size))
-	return nil
-}
-
-func (b *blobDownload) Run(ctx context.Context, requestURL *url.URL, opts *RegistryOptions) {
-	b.err = b.run(ctx, requestURL, opts)
-}
-
-func (b *blobDownload) run(ctx context.Context, requestURL *url.URL, opts *RegistryOptions) error {
-	defer blobDownloadManager.Delete(b.Digest)
-
-	ctx, b.CancelFunc = context.WithCancel(ctx)
-
-	file, err := os.OpenFile(b.Name+"-partial", os.O_CREATE|os.O_RDWR, 0644)
-	if err != nil {
-		return err
-	}
-	defer file.Close()
-
-	file.Truncate(b.Total)
-
-	g, inner := errgroup.WithContext(ctx)
-	g.SetLimit(numDownloadParts)
-	for i := range b.Parts {
-		part := b.Parts[i]
-		if part.Completed == part.Size {
-			continue
-		}
-
-		i := i
-		g.Go(func() error {
-			for try := 0; try < maxRetries; try++ {
-				w := io.NewOffsetWriter(file, part.StartsAt())
-				err := b.downloadChunk(inner, requestURL, w, part, opts)
-				switch {
-				case errors.Is(err, context.Canceled):
-					return err
-				case err != nil:
-					log.Printf("%s part %d attempt %d failed: %v, retrying", b.Digest[7:19], i, try, err)
-					continue
-				default:
-					return nil
-				}
-			}
-
-			return errors.New("max retries exceeded")
-		})
-	}
-
-	if err := g.Wait(); err != nil {
-		return err
-	}
-
-	// explicitly close the file so we can rename it
-	if err := file.Close(); err != nil {
-		return err
-	}
-
-	for i := range b.Parts {
-		if err := os.Remove(file.Name() + "-" + strconv.Itoa(i)); err != nil {
-			return err
-		}
-	}
-
-	if err := os.Rename(file.Name(), b.Name); err != nil {
-		return err
-	}
-
-	b.done = true
-	return nil
-}
-
-func (b *blobDownload) downloadChunk(ctx context.Context, requestURL *url.URL, w io.Writer, part *blobDownloadPart, opts *RegistryOptions) error {
-	headers := make(http.Header)
-	headers.Set("Range", fmt.Sprintf("bytes=%d-%d", part.StartsAt(), part.StopsAt()-1))
-	resp, err := makeRequest(ctx, "GET", requestURL, headers, nil, opts)
-	if err != nil {
-		return err
-	}
-	defer resp.Body.Close()
-
-	n, err := io.Copy(w, io.TeeReader(resp.Body, b))
-	if err != nil && !errors.Is(err, context.Canceled) {
-		// rollback progress
-		b.Completed.Add(-n)
-		return err
-	}
-
-	part.Completed += n
-	if err := b.writePart(part.Name(), part); err != nil {
-		return err
-	}
-
-	// return nil or context.Canceled
-	return err
-}
-
-func (b *blobDownload) newPart(offset, size int64) error {
-	part := blobDownloadPart{blobDownload: b, Offset: offset, Size: size, N: len(b.Parts)}
-	if err := b.writePart(part.Name(), &part); err != nil {
-		return err
-	}
-
-	b.Parts = append(b.Parts, &part)
-	return nil
-}
-
-func (b *blobDownload) readPart(partName string) (*blobDownloadPart, error) {
-	var part blobDownloadPart
-	partFile, err := os.Open(partName)
-	if err != nil {
-		return nil, err
-	}
-	defer partFile.Close()
-
-	if err := json.NewDecoder(partFile).Decode(&part); err != nil {
-		return nil, err
-	}
-
-	part.blobDownload = b
-	return &part, nil
-}
-
-func (b *blobDownload) writePart(partName string, part *blobDownloadPart) error {
-	partFile, err := os.OpenFile(partName, os.O_CREATE|os.O_RDWR|os.O_TRUNC, 0644)
-	if err != nil {
-		return err
-	}
-	defer partFile.Close()
-
-	return json.NewEncoder(partFile).Encode(part)
-}
-
-func (b *blobDownload) Write(p []byte) (n int, err error) {
-	n = len(p)
-	b.Completed.Add(int64(n))
-	return n, nil
-}
-
-func (b *blobDownload) acquire() {
-	b.references.Add(1)
-}
-
-func (b *blobDownload) release() {
-	if b.references.Add(-1) == 0 {
-		b.CancelFunc()
-	}
-}
-
-func (b *blobDownload) Wait(ctx context.Context, fn func(api.ProgressResponse)) error {
-	b.acquire()
-	defer b.release()
-
-	ticker := time.NewTicker(60 * time.Millisecond)
-	for {
-		select {
-		case <-ticker.C:
-		case <-ctx.Done():
-			return ctx.Err()
-		}
-
-		fn(api.ProgressResponse{
-			Status:    fmt.Sprintf("downloading %s", b.Digest),
-			Digest:    b.Digest,
-			Total:     b.Total,
-			Completed: b.Completed.Load(),
-		})
-
-		if b.done || b.err != nil {
-			return b.err
-		}
-	}
-}
+var inProgress sync.Map // map of digests currently being downloaded to their current download progress

 type downloadOpts struct {
 	mp      ModelPath
 	digest  string
 	regOpts *RegistryOptions
 	fn      func(api.ProgressResponse)
+	retry   int // track the number of retries on this download
 }

-const maxRetries = 3
+const maxRetry = 3

 // downloadBlob downloads a blob from the registry and stores it in the blobs directory
 func downloadBlob(ctx context.Context, opts downloadOpts) error {
@@ -313,14 +42,9 @@ func downloadBlob(ctx context.Context, opts downloadOpts) error {
 		return err
 	}

-	fi, err := os.Stat(fp)
-	switch {
-	case errors.Is(err, os.ErrNotExist):
-	case err != nil:
-		return err
-	default:
+	if fi, _ := os.Stat(fp); fi != nil {
+		// we already have the file, so return
 		opts.fn(api.ProgressResponse{
-			Status:    fmt.Sprintf("downloading %s", opts.digest),
 			Digest:    opts.digest,
 			Total:     fi.Size(),
 			Completed: fi.Size(),
@@ -329,18 +53,185 @@ func downloadBlob(ctx context.Context, opts downloadOpts) error {
 		return nil
 	}

-	data, ok := blobDownloadManager.LoadOrStore(opts.digest, &blobDownload{Name: fp, Digest: opts.digest})
-	download := data.(*blobDownload)
-	if !ok {
-		requestURL := opts.mp.BaseURL()
-		requestURL = requestURL.JoinPath("v2", opts.mp.GetNamespaceRepository(), "blobs", opts.digest)
-		if err := download.Prepare(ctx, requestURL, opts.regOpts); err != nil {
-			blobDownloadManager.Delete(opts.digest)
-			return err
-		}
-
-		go download.Run(context.Background(), requestURL, opts.regOpts)
+	fileDownload := &FileDownload{
+		Digest:    opts.digest,
+		FilePath:  fp,
+		Total:     1, // dummy value to indicate that we don't know the total size yet
+		Completed: 0,
 	}

-	return download.Wait(ctx, opts.fn)
+	_, downloading := inProgress.LoadOrStore(opts.digest, fileDownload)
+	if downloading {
+		// this is another client requesting the server to download the same blob concurrently
+		return monitorDownload(ctx, opts, fileDownload)
+	}
+	if err := doDownload(ctx, opts, fileDownload); err != nil {
+		if errors.Is(err, errDownload) && opts.retry < maxRetry {
+			opts.retry++
+			log.Print(err)
+			log.Printf("retrying download of %s", opts.digest)
+			return downloadBlob(ctx, opts)
+		}
+		return err
+	}
+	return nil
+}
+
+var downloadMu sync.Mutex // mutex to check to resume a download while monitoring
+
+// monitorDownload monitors the download progress of a blob and resumes it if it is interrupted
+func monitorDownload(ctx context.Context, opts downloadOpts, f *FileDownload) error {
+	tick := time.NewTicker(time.Second)
+	for range tick.C {
+		done, resume, err := func() (bool, bool, error) {
+			downloadMu.Lock()
+			defer downloadMu.Unlock()
+			val, downloading := inProgress.Load(f.Digest)
+			if !downloading {
+				// check once again if the download is complete
+				if fi, _ := os.Stat(f.FilePath); fi != nil {
+					// successful download while monitoring
+					opts.fn(api.ProgressResponse{
+						Digest:    f.Digest,
+						Total:     fi.Size(),
+						Completed: fi.Size(),
+					})
+					return true, false, nil
+				}
+				// resume the download
+				inProgress.Store(f.Digest, f) // store the file download again to claim the resume
+				return false, true, nil
+			}
+			f, ok := val.(*FileDownload)
+			if !ok {
+				return false, false, fmt.Errorf("invalid type for in progress download: %T", val)
+			}
+			opts.fn(api.ProgressResponse{
+				Status:    fmt.Sprintf("downloading %s", f.Digest),
+				Digest:    f.Digest,
+				Total:     f.Total,
+				Completed: f.Completed,
+			})
+			return false, false, nil
+		}()
+		if err != nil {
+			return err
+		}
+		if done {
+			// done downloading
+			return nil
+		}
+		if resume {
+			return doDownload(ctx, opts, f)
+		}
+	}
+	return nil
+}
+
+var (
+	chunkSize   int64 = 1024 * 1024 // 1 MiB in bytes
+	errDownload       = fmt.Errorf("download failed")
+)
+
+// doDownload downloads a blob from the registry and stores it in the blobs directory
+func doDownload(ctx context.Context, opts downloadOpts, f *FileDownload) error {
+	defer inProgress.Delete(f.Digest)
+	var size int64
+
+	fi, err := os.Stat(f.FilePath + "-partial")
+	switch {
+	case errors.Is(err, os.ErrNotExist):
+		// noop, file doesn't exist so create it
+	case err != nil:
+		return fmt.Errorf("stat: %w", err)
+	default:
+		size = fi.Size()
+		// Ensure the size is divisible by the chunk size by removing excess bytes
+		size -= size % chunkSize
+
+		err := os.Truncate(f.FilePath+"-partial", size)
+		if err != nil {
+			return fmt.Errorf("truncate: %w", err)
+		}
+	}
+
+	requestURL := opts.mp.BaseURL()
+	requestURL = requestURL.JoinPath("v2", opts.mp.GetNamespaceRepository(), "blobs", f.Digest)
+
+	headers := make(http.Header)
+	headers.Set("Range", fmt.Sprintf("bytes=%d-", size))
+
+	resp, err := makeRequest(ctx, "GET", requestURL, headers, nil, opts.regOpts)
+	if err != nil {
+		log.Printf("couldn't download blob: %v", err)
+		return fmt.Errorf("%w: %w", errDownload, err)
+	}
+	defer resp.Body.Close()
+
+	if resp.StatusCode >= http.StatusBadRequest {
+		body, _ := io.ReadAll(resp.Body)
+		return fmt.Errorf("%w: on download registry responded with code %d: %v", errDownload, resp.StatusCode, string(body))
+	}
+
+	err = os.MkdirAll(filepath.Dir(f.FilePath), 0o700)
+	if err != nil {
+		return fmt.Errorf("make blobs directory: %w", err)
+	}
+
+	remaining, _ := strconv.ParseInt(resp.Header.Get("Content-Length"), 10, 64)
+	f.Completed = size
+	f.Total = remaining + f.Completed
+
+	inProgress.Store(f.Digest, f)
+
+	out, err := os.OpenFile(f.FilePath+"-partial", os.O_CREATE|os.O_APPEND|os.O_WRONLY, 0o644)
+	if err != nil {
+		return fmt.Errorf("open file: %w", err)
+	}
+	defer out.Close()
+outerLoop:
+	for {
+		select {
+		case <-ctx.Done():
+			// handle client request cancellation
+			inProgress.Delete(f.Digest)
+			return nil
+		default:
+			opts.fn(api.ProgressResponse{
+				Status:    fmt.Sprintf("downloading %s", f.Digest),
+				Digest:    f.Digest,
+				Total:     f.Total,
+				Completed: f.Completed,
+			})
+
+			if f.Completed >= f.Total {
+				if err := out.Close(); err != nil {
+					return err
+				}
+
+				if err := os.Rename(f.FilePath+"-partial", f.FilePath); err != nil {
+					opts.fn(api.ProgressResponse{
+						Status:    fmt.Sprintf("error renaming file: %v", err),
+						Digest:    f.Digest,
+						Total:     f.Total,
+						Completed: f.Completed,
+					})
+					return err
+				}
+
+				break outerLoop
+			}
+		}
+
+		n, err := io.CopyN(out, resp.Body, chunkSize)
+		if err != nil && !errors.Is(err, io.EOF) {
+			return fmt.Errorf("%w: %w", errDownload, err)
+		}
+		f.Completed += n
+
+		inProgress.Store(f.Digest, f)
+	}
+
+	log.Printf("success getting %s\n", f.Digest)
+	return nil
 }
--- a/server/images.go
+++ b/server/images.go
@@ -30,6 +30,8 @@ import (
 	"github.com/jmorganca/ollama/version"
 )

+const MaxRetries = 3
+
 type RegistryOptions struct {
 	Insecure bool
 	Username string
@@ -52,6 +54,54 @@ type Model struct {
 	Embeddings    []vector.Embedding
 }

+func (m *Model) ChatPrompt(messages []api.Message) (string, error) {
+	tmpl, err := template.New("").Parse(m.Template)
+	if err != nil {
+		return "", err
+	}
+
+	var vars struct {
+		System string
+		Prompt string
+		First  bool
+	}
+
+	vars.First = true
+
+	var sb strings.Builder
+	flush := func() {
+		tmpl.Execute(&sb, vars)
+		vars.System = ""
+		vars.Prompt = ""
+	}
+
+	// build the chat history from messages
+	for _, m := range messages {
+		if m.Role == "system" {
+			if vars.System != "" {
+				flush()
+			}
+			vars.System = m.Content
+		}
+
+		if m.Role == "user" {
+			if vars.Prompt != "" {
+				flush()
+			}
+			vars.Prompt = m.Content
+		}
+
+		if m.Role == "assistant" {
+			flush()
+			sb.Write([]byte(m.Content))
+		}
+	}
+
+	flush()
+
+	return sb.String(), nil
+}
+
 func (m *Model) Prompt(request api.GenerateRequest, embedding string) (string, error) {
 	t := m.Template
 	if request.Template != "" {
@@ -372,7 +422,6 @@ func CreateModel(ctx context.Context, workDir, name string, path string, fn func
 			}

 			if mf != nil {
-				fn(api.ProgressResponse{Status: "reading model metadata"})
 				sourceBlobPath, err := GetBlobsPath(mf.Config.Digest)
 				if err != nil {
 					return err
@@ -830,7 +879,7 @@ func formatParams(params map[string][]string) (map[string]interface{}, error) {
 						return nil, fmt.Errorf("invalid int value %s", vals)
 					}

-					out[key] = intVal
+					out[key] = int(intVal)
 				case reflect.Bool:
 					boolVal, err := strconv.ParseBool(vals[0])
 					if err != nil {
@@ -1416,7 +1465,7 @@ func checkBlobExistence(ctx context.Context, mp ModelPath, digest string, regOpt

 func makeRequestWithRetry(ctx context.Context, method string, requestURL *url.URL, headers http.Header, body io.ReadSeeker, regOpts *RegistryOptions) (*http.Response, error) {
 	var status string
-	for try := 0; try < maxRetries; try++ {
+	for try := 0; try < MaxRetries; try++ {
 		resp, err := makeRequest(ctx, method, requestURL, headers, body, regOpts)
 		if err != nil {
 			log.Printf("couldn't start upload: %v", err)
@@ -1486,14 +1535,13 @@ func makeRequest(ctx context.Context, method string, requestURL *url.URL, header
 		req.ContentLength = contentLength
 	}

-	proxyURL, err := http.ProxyFromEnvironment(req)
-	if err != nil {
-		return nil, err
-	}
-
-	client := http.Client{
-		Transport: &http.Transport{
-			Proxy: http.ProxyURL(proxyURL),
+	client := &http.Client{
+		CheckRedirect: func(req *http.Request, via []*http.Request) error {
+			if len(via) >= 10 {
+				return fmt.Errorf("too many redirects")
+			}
+			log.Printf("redirected to: %s\n", req.URL)
+			return nil
 		},
 	}

--- a/server/routes.go
+++ b/server/routes.go
@@ -3,10 +3,8 @@ package server
 import (
 	"context"
 	"encoding/json"
-	"errors"
 	"fmt"
 	"io"
-	"io/fs"
 	"log"
 	"net"
 	"net/http"
@@ -68,6 +66,7 @@ func load(ctx context.Context, workDir string, model *Model, reqOpts map[string]
 	}

 	if err := opts.FromMap(reqOpts); err != nil {
+		log.Printf("could not merge model options: %v", err)
 		return err
 	}

@@ -157,6 +156,54 @@ func load(ctx context.Context, workDir string, model *Model, reqOpts map[string]
 	return nil
 }

+func ChatModelHandler(c *gin.Context) {
+	loaded.mu.Lock()
+	defer loaded.mu.Unlock()
+
+	var req api.ChatRequest
+	if err := c.ShouldBindJSON(&req); err != nil {
+		c.JSON(http.StatusBadRequest, gin.H{"error": err.Error()})
+		return
+	}
+
+	model, err := GetModel(req.Model)
+	if err != nil {
+		c.JSON(http.StatusBadRequest, gin.H{"error": err.Error()})
+		return
+	}
+
+	prompt, err := model.ChatPrompt(req.Messages)
+	if err != nil {
+		c.JSON(http.StatusInternalServerError, gin.H{"error": err.Error()})
+		return
+	}
+
+	var response string
+	fn := func(r api.GenerateResponse) {
+		response += r.Response
+	}
+
+	workDir := c.GetString("workDir")
+	if err := load(c.Request.Context(), workDir, model, nil, defaultSessionDuration); err != nil {
+		c.JSON(http.StatusInternalServerError, gin.H{"error": err.Error()})
+		return
+	}
+
+	fmt.Println(prompt)
+
+	if err := loaded.llm.Predict(c.Request.Context(), []int{}, prompt, fn); err != nil {
+		c.JSON(http.StatusBadRequest, gin.H{"error": err.Error()})
+	}
+
+	c.JSON(http.StatusOK, api.ChatResponse{
+		Message: api.Message{
+			Role:    "assistant",
+			Content: response,
+		},
+		CreatedAt: time.Now().UTC(),
+	})
+}
+
 func GenerateHandler(c *gin.Context) {
 	loaded.mu.Lock()
 	defer loaded.mu.Unlock()
@@ -171,11 +218,6 @@ func GenerateHandler(c *gin.Context) {

 	model, err := GetModel(req.Model)
 	if err != nil {
-		var pErr *fs.PathError
-		if errors.As(err, &pErr) {
-			c.JSON(http.StatusNotFound, gin.H{"error": fmt.Sprintf("model '%s' not found, try pulling it first", req.Model)})
-			return
-		}
 		c.JSON(http.StatusBadRequest, gin.H{"error": err.Error()})
 		return
 	}
@@ -185,10 +227,6 @@ func GenerateHandler(c *gin.Context) {
 	// TODO: set this duration from the request if specified
 	sessionDuration := defaultSessionDuration
 	if err := load(c.Request.Context(), workDir, model, req.Options, sessionDuration); err != nil {
-		if errors.Is(err, api.ErrInvalidOpts) {
-			c.JSON(http.StatusBadRequest, gin.H{"error": err.Error()})
-			return
-		}
 		c.JSON(http.StatusInternalServerError, gin.H{"error": err.Error()})
 		return
 	}
@@ -243,23 +281,6 @@ func GenerateHandler(c *gin.Context) {
 		}
 	}()

-	if req.Stream != nil && !*req.Stream {
-		var response api.GenerateResponse
-		generated := ""
-		for resp := range ch {
-			if r, ok := resp.(api.GenerateResponse); ok {
-				generated += r.Response
-				response = r
-			} else {
-				c.JSON(http.StatusInternalServerError, gin.H{"error": err.Error()})
-				return
-			}
-		}
-		response.Response = generated
-		c.JSON(http.StatusOK, response)
-		return
-	}
-
 	streamResponse(c, ch)
 }

@@ -329,11 +350,6 @@ func PullModelHandler(c *gin.Context) {
 		}
 	}()

-	if req.Stream != nil && !*req.Stream {
-		waitForStream(c, ch)
-		return
-	}
-
 	streamResponse(c, ch)
 }

@@ -361,11 +377,6 @@ func PushModelHandler(c *gin.Context) {
 		}
 	}()

-	if req.Stream != nil && !*req.Stream {
-		waitForStream(c, ch)
-		return
-	}
-
 	streamResponse(c, ch)
 }

@@ -393,11 +404,6 @@ func CreateModelHandler(c *gin.Context) {
 		}
 	}()

-	if req.Stream != nil && !*req.Stream {
-		waitForStream(c, ch)
-		return
-	}
-
 	streamResponse(c, ch)
 }

@@ -594,6 +600,7 @@ func Serve(ln net.Listener, allowOrigins []string) error {
 		},
 	)

+	r.POST("/api/chat", ChatModelHandler)
 	r.POST("/api/pull", PullModelHandler)
 	r.POST("/api/generate", GenerateHandler)
 	r.POST("/api/embeddings", EmbeddingHandler)
@@ -638,31 +645,6 @@ func Serve(ln net.Listener, allowOrigins []string) error {
 	return s.Serve(ln)
 }

-func waitForStream(c *gin.Context, ch chan interface{}) {
-	c.Header("Content-Type", "application/json")
-	for resp := range ch {
-		switch r := resp.(type) {
-		case api.ProgressResponse:
-			if r.Status == "success" {
-				c.JSON(http.StatusOK, r)
-				return
-			}
-		case gin.H:
-			if errorMsg, ok := r["error"].(string); ok {
-				c.JSON(http.StatusInternalServerError, gin.H{"error": errorMsg})
-				return
-			} else {
-				c.JSON(http.StatusInternalServerError, gin.H{"error": "unexpected error format in progress response"})
-				return
-			}
-		default:
-			c.JSON(http.StatusInternalServerError, gin.H{"error": "unexpected progress response"})
-			return
-		}
-	}
-	c.JSON(http.StatusInternalServerError, gin.H{"error": "unexpected end of progress response"})
-}
-
 func streamResponse(c *gin.Context, ch chan any) {
 	c.Header("Content-Type", "application/x-ndjson")
 	c.Stream(func(w io.Writer) bool {
--- a/server/upload.go
+++ b/server/upload.go
@@ -10,7 +10,6 @@ import (
 	"net/url"
 	"os"
 	"strconv"
-	"sync"

 	"github.com/jmorganca/ollama/api"
 )
@@ -139,7 +138,7 @@ func uploadBlobChunk(ctx context.Context, method string, requestURL *url.URL, r
 		headers.Set("Content-Range", fmt.Sprintf("%d-%d", offset, offset+sectionReader.Size()-1))
 	}

-	for try := 0; try < maxRetries; try++ {
+	for try := 0; try < MaxRetries; try++ {
 		resp, err := makeRequest(ctx, method, requestURL, headers, io.TeeReader(sectionReader, pw), opts)
 		if err != nil && !errors.Is(err, io.EOF) {
 			return nil, err
@@ -192,13 +191,9 @@ type ProgressWriter struct {
 	completed int64
 	total     int64
 	fn        func(api.ProgressResponse)
-	mu        sync.Mutex
 }

 func (pw *ProgressWriter) Write(b []byte) (int, error) {
-	pw.mu.Lock()
-	defer pw.mu.Unlock()
-
 	n := len(b)
 	pw.bucket += int64(n)