The primary is a caching mirror of the same two files, so an install
resolves its gallery from infrastructure the project controls. The GitHub
URI stays as a mirror, so behaviour is unchanged whenever the primary is
unreachable.
Docs and agent guides that either state the shipped defaults or hand out a
copy-pasteable gallery list are updated to match, so following them no
longer silently demotes an install off the new primary.
Assisted-by: Claude:claude-opus-5 [go vet] [go test] [golangci-lint]
Co-authored-by: Ettore Di Giacinto <mudler@localai.io>
* feat(version): include OS and arch in the outbound User-Agent
Registries and galleries already receive LocalAI/<version>; adding the
platform follows ordinary client convention and discloses nothing a
registry cannot infer from the manifest it is asked for.
Updates the User-Agent note in docs/content/getting-started/models.md,
which documented the old format.
Assisted-by: Claude:claude-opus-5 [go vet] [go test]
* feat(downloader): identify LocalAI on outbound requests
pkg/oci has always sent a User-Agent; the downloader sent none, so gallery
reads, model-file downloads, resume probes, content-length probes and the
HuggingFace safety scan all went out as a bare Go HTTP client, unattributable
to LocalAI by the hosts serving them.
HuggingFaceScan moves off the client's Get shorthand to an explicit request
for the same reason — the shorthand gives no place to hang a header.
Extends the User-Agent note in docs/content/getting-started/models.md, which
claimed the header was sent only to Ollama and OCI registries.
Assisted-by: Claude:claude-opus-5 [go vet] [go test]
* feat(gallery): add a mirrors list to gallery configuration
Mirrors are an availability fallback, tried in order only after the primary
URL fails. omitempty keeps existing configurations byte-identical.
The slice makes config.Gallery non-comparable with ==, which broke the two
slices.Equal callers in the runtime settings registry. Replace them with an
explicit Gallery.Equal / GalleriesEqual so a gallery list that differs from
the baseline only by its mirrors still counts as env/CLI-set. Equal compares
the Verification block by value; == compared it by pointer identity, which
called two structurally identical policies different.
Assisted-by: Claude:claude-opus-5 [go vet] [go test]
* fix(downloader): treat an HTTP error status as a failed read
ReadWithCallback handed the response body to its callback whatever the
status was, so a 404 page or a 502 from a CDN arrived as if it were a
gallery index or a model config: it parsed to nothing, got cached for an
hour, and no caller could tell the source had been down. DownloadFile has
always checked the status; this path never did.
Mirror fallback depends on it — a source that answers with an error page
has to count as unreachable, or the next candidate is never tried.
Assisted-by: Claude:claude-opus-5 [go vet] [go test]
* feat(gallery): fall back to mirrors when the primary source fails
Candidates are tried primary-first with a bounded timeout each, and a
source that just failed is skipped for a cooldown so a dead host is not
re-dialled on every listing. When every candidate is in cooldown they are
all tried anyway: refusing to serve a gallery we might be able to reach is
worse than one slow request.
The one-hour index cache is untouched and stays keyed on the gallery's own
identity, so a mirror-served fetch fills the entry the primary would have.
No SSRF validation is applied to the candidates. validateGalleryConfigURL
guards GetGalleryConfigFromURL because that URL arrives in a request body;
mirrors come from the operator's gallery configuration, the same place the
primary has always come from, and the index fetch has never validated the
primary. Validating mirrors while the primary goes unchecked would buy
nothing and would break the deployment mirrors exist for — an index served
from a host on the LAN.
Assisted-by: Claude:claude-opus-5 [go vet] [go test]
* fix(gallery): loosen the mirror fetch timeout and stop blaming the caller
The downloader only ever bounded response headers, never the body, so the
per-attempt deadline added with mirror fallback was the first whole-transfer
timeout this path has had. At 30s the default 2.2 MB index demanded ~75 KB/s
sustained: a rural-DSL, mobile or satellite user who used to wait 60s and
succeed would now fail, and then eat a 10-minute cooldown on a source that
was perfectly healthy. Raised to 120s (~19 KB/s), which no link that could
go on to download a model will miss, and made it a var so a test can shorten
it and prove a hanging candidate is actually abandoned.
Caller cancellation is no longer recorded as a failure of the source.
Unreachable today since getGalleryElements passes context.Background(), but
once a request context is wired through, a browser disconnect would have
blackholed every candidate for ten minutes over something the sources had
no part in.
Also document that mirrors do not cover a .ref gallery URL: the reference is
resolved before mirrors are considered, so a .ref that cannot be fetched
fails the gallery outright. Routing .ref resolution through the candidate
list needs a per-candidate resolve-and-fetch and a decision about cache
identity, which is more than this change should carry.
Assisted-by: Claude:claude-opus-5 [go vet] [go test]
* feat(gallery): serve the last known good index when everything is offline
A successful fetch is cached alongside the models directory and served when
no source is reachable, so an offline or airgapped machine can still list
its gallery. Entries may be stale in that state, and the fallback is logged.
The copy is deliberately kept out of the models directory, where a <name>.yaml
file is read as an installed model's configuration, and is named after a digest
of the gallery URL so the model and backend galleries cannot collide. Writing
it is best effort: a read-only or full disk must not fail a fetch that
otherwise succeeded.
Also corrects the mirror scheme list in the docs: the HuggingFace prefixes are
huggingface://, hf:// and hf.co/, not huggingface:.
Assisted-by: Claude:claude-opus-5 [go vet] [go test]
* fix(gallery): only cache a response that is really a gallery index
The last known good copy was written on any 2xx, before anything looked
at the bytes: the parse only happens later, in getGalleryElements. A
captive portal, a corporate proxy or a CDN error page all answer HTTP 200
with HTML, so any of them could overwrite a good copy. The listing fails
then and there, and the next offline start — the one case this cache
exists for — serves the interception page instead of the gallery it
already had.
Probe the body before persisting it: unmarshal into a []any and keep the
older copy unless the result is a non-empty sequence. An empty document
is rejected too. It parses fine, so a parse-only check would still let a
blank response replace a populated index with one that lists nothing,
which from the user's side is the same outage; and an empty index is
worth nothing offline, so there is no case where caching it beats keeping
what came before. The live body is still returned to the caller — the
probe gates persistence only, and getGalleryElements remains the thing
that reports a real parse failure.
Also in this pass:
- The empty-basePath guard only caught exact "". galleryCachePath(".")
and galleryCachePath("models") still resolved the cache sibling against
the process working directory, which is what the guard was written to
prevent. Reject any non-absolute base.
- The docs claimed the offline cache "applies to every gallery, with or
without mirrors". Not true for a .ref URL: the reference is resolved
before the cache is consulted, so a .ref gallery fails offline even
after a successful earlier fetch, and the cache file it writes can
never be read. Extend the .ref warning and qualify the sentence.
- pkg/oci's UserAgent comment never mentioned the platform component
added earlier on this branch.
- resetGalleryFailures and expireGalleryFailure had no non-test callers;
move them into the test file.
- The all-candidates-failed error reported len(attempt), so a three
mirror gallery with two sources in cooldown said "all 1 source(s)
failed" — which reads as a misconfiguration. Report how many were
configured and how many were skipped.
- Give the package's tests their own TMPDIR. The cache is a sibling of
the models directory, which is right in production, but specs that
build a models directory directly under /tmp made the sibling resolve
to /tmp/cache and left it behind after every run.
Assisted-by: Claude:claude-opus-5 [go vet] [go test]
* fix(gallery): convert the new tests to Ginkgo and clear the lint gate
.agents/coding-style.md requires Ginkgo v2 + Gomega for every Go test and
has forbidigo enforce it; the stdlib-style tests still in the tree are tech
debt, not a pattern. Every test file this branch added was written in the
forbidden style, which is what turned CI red.
Convert all five of them. internal had no suite bootstrap, so add one;
core/config, core/gallery and pkg/downloader already have theirs and are
reused, so no package mixes styles. pkg/downloader/useragent_test.go and
read_status_test.go were not in CI's forbidigo list but used the same
forbidden calls, so they are converted too.
The one conversion with a trap in it is core/gallery. Go's t.TempDir()
yields $TMPDIR/<TestName>NNNN/001, so the gallery cache — a sibling of the
models directory — was isolated per test. GinkgoT().TempDir() yields a flat
$TMPDIR/ginkgoNNNN, which would put every spec's cache in one shared
directory and break the specs that count files in it. tempModelsDir()
restores the original isolation.
Also make the deliberate cleanup-path ignores explicit with `_ =`, drop the
gallery cache directory to 0750 (nothing outside the server's own user and
group reads it), and justify the cache read with a #nosec G304 comment in
the form already used elsewhere in the tree: the path is a hex sha256 under
a fixed directory with a non-absolute base already rejected, so no
caller-supplied text reaches it.
Re-ran the mutations these specs were verified against — dropping the
platform suffix from UserAgent, making Gallery.Equal ignore Mirrors and
ignore Name, removing persistGalleryIndex's validity probe, removing the
!filepath.IsAbs guard, not skipping a cooled-down candidate, and dropping
the per-attempt timeout. All seven still fail the converted specs.
Assisted-by: Claude:claude-opus-5 [go vet] [go test] [golangci-lint] [gosec]
---------
Co-authored-by: Ettore Di Giacinto <mudler@localai.io>
A container engine started from a socket-activated system unit leaks a bare
LISTEN_PID into every container it spawns, with no matching LISTEN_FDS. LocalAI
read that as a malformed activation attempt and refused to start:
ERROR Error running the application error=loading systemd socket
activation listeners: invalid LISTEN_FDS ""
systemd's own sd_listen_fds() treats either variable being absent as "not
activated" rather than as an error, so do the same and fall back to ordinary
--address binding. A value that is present but malformed is still rejected, so
a real activation attempt cannot silently bind the wrong socket.
Fixes#11390
Assisted-by: Claude:claude-opus-5 [golangci-lint]
Signed-off-by: Ettore Di Giacinto <mudler@localai.io>
Co-authored-by: Ettore Di Giacinto <mudler@localai.io>
* fix(react-ui): restore 3D Studio results and history
Keep large conditioning-image payloads out of the rendered request panel so the generated viewer can mount reliably. Accept clipboard images and synchronize 3D history consumers so new results appear in Studio without a reload.
Cover clipboard input, bounded request rendering, result display, and cross-view history synchronization with Playwright.
Assisted-by: Codex:gpt-5 Playwright
* perf(react-ui): idle the 3D viewport when still
Limit auto-rotate rendering to 30 FPS and stop scheduling frames when rotation is disabled. Resize, view controls, and pointer input invalidate the still frame on demand.
Assisted-by: Codex:gpt-5 Playwright
The backend could configure four of the engine's knobs (block size, KV block
count, max sequence length, max concurrent sequences) out of a config surface
that is considerably larger. Speculative decoding, prefix caching, the
chunked-prefill token budget, the scheduling policy and the external KV
connector were reachable from vllm.cpp's own HTTP server and from nothing
LocalAI could write in a model config.
Config now goes through `engine_args:`, the same map the vLLM and SGLang
backends take, with keys spelled as vLLM's own CLI flags so a speculative_config
or kv_transfer_config block written for vLLM works verbatim. The legacy
`options:` list keeps working and reads every key too; engine_args wins where
both set one. Unknown keys are logged and ignored rather than fatal: the field
is shared with the other engines, so a config carrying their knobs must not take
the model down.
Two details worth knowing:
`enable_prefix_caching: false` maps to the ABI tri-state force-OFF (2), not 0.
0 means "let the model capability decide" and dense architectures default the
cache on, so collapsing the two would silently enable it against an explicit
false. enable_jump_forward (ABI v10) shares the encoding, deferring to
VT_ENABLE_JUMP_FORWARD instead of to the model.
The importer probes config.json on a vllm-cpp import and writes
speculative_config: {method: mtp} when the checkpoint declares an MTP head, the
safetensors analogue of the llama-cpp importer's GGUF probe. DFlash draft repos
are refused with a warning instead, since a drafter cannot serve alone and the
pairing is not derivable from either repo. The draft path is resolved against
LocalAI's model directory, because the engine only looks in a directory holding
config.json or in the HF cache and never downloads: the repo-id spelling the
vLLM docs teach used to die deep in the load with "draft checkpoint not found".
docs/content/features/text-generation.md gains a vllm.cpp section covering the
engine_args table, all three speculative methods, LMCache and the legacy list.
The backend had no documentation page before.
This replaces a branch that had gone stale behind master and carried its own
route to ABI v10, which #11386 has since landed in minimal form. Rebased onto
that as a single commit rather than replaying the intermediate steps, whose
ABI v9 mirrors no longer make sense against master's pin. The Darwin build
fixes for Apple Clang's gnu-folding-constant diagnostic on C++, Objective-C and
Objective-C++, originally authored by localai-org-maint-bot, are folded in here.
Verified: `make abi-check` agrees at v10; unit specs, core/config and
core/gallery/importers green; and the full e2e passes in 1330s against a CPU
libvllm.so reporting ABI v10 with Qwen_Qwen3.5-0.8B-Q4_K_M.gguf (load, blocking
completion, streaming, chat and tool calls).
Assisted-by: Claude:claude-fable-5 golangci-lint
Signed-off-by: Ettore Di Giacinto <mudler@localai.io>
Co-authored-by: Ettore Di Giacinto <mudler@localai.io>
Add the MTP and plain Q4_K_M GGUF builds with their shared vision projector so LocalAI users can select accelerated or fallback llama.cpp inference.
Assisted-by: Codex:gpt-5
Co-authored-by: localai-org-maint-bot <306269227+localai-org-maint-bot@users.noreply.github.com>
Add Q4_K_M and Q8_0 builds of the popular refusal-removed Qwen3.5 9B fine-tune, including its multimodal projector.
Assisted-by: Codex:gpt-5 [web]
Co-authored-by: localai-org-maint-bot <306269227+localai-org-maint-bot@users.noreply.github.com>
Switching from Backend Traces back to API Traces crashed the page
with "can't access property status, e.response is undefined" (#11376).
The API table briefly renders the previous tab's backend rows while the
refetch effect is still pending, and those rows carry no `response`
envelope. The status column dereferenced it unguarded. Render a neutral
placeholder instead of throwing, and cover the tab-switch scenario with
a regression spec.
Assisted-by: opencode:big-pickle
Signed-off-by: Nandana Dileep <110280757+nandanadileep@users.noreply.github.com>
The Go bindings mirror vllm.h by hand and refuse a library whose
vllm_abi_version differs from what they were written against. Two
automated pin bumps (#11174, #11352) moved VLLM_CPP_VERSION onto engines
declaring ABI v10 while govllmcpp.go still mirrored v5, so every
vllm-cpp image built since then panics at startup on every platform:
panic: vllm-cpp: ABI mismatch: library reports v10, backend built against v5
Grow both PODs to the v10 layout: vllm_model_params gains
speculative_config, enable_prefix_caching, max_num_batched_tokens,
scheduling_policy, kv_transfer_config and enable_jump_forward (88 bytes),
vllm_sampling_params gains the v8 logits-processor pair (136 bytes). The
offsets in the specs come from offsetof() against the pinned header. All
of the new fields are inert when zeroed, so the engine behaves exactly as
it did under v5; the backend sets none of them.
Nothing cross-checked the two files, which is why a blind pin bump could
ship a backend that cannot load. The library build now runs abi-check
first: it compares VLLM_ABI_VERSION in the fetched header against
abiVersion in govllmcpp.go and fails the build naming both, instead of
leaving the mismatch for a user's runtime.
Fixes#11379
Assisted-by: Claude:claude-fable-5 golangci-lint
Signed-off-by: Ettore Di Giacinto <mudler@localai.io>
Co-authored-by: Ettore Di Giacinto <mudler@localai.io>
Recover parser panics at metadata boundaries, skip unneeded remote arrays, and use the parser's overflow-hardened release. Keep detached gallery workers and CrispASR probes from terminating their processes on malformed GGUF input. Disable startup warming in the provided Compose files as an operational fallback.
Assisted-by: Codex:gpt-5
Signed-off-by: Richard Palethorpe <io@richiejp.com>
docs(blog): cover the terminal agent, and fix the counts in the intro
The 4.8 post never mentions that `local-ai chat` stopped being a REPL
and became an agent (#11291): the nib harness compiled into the binary,
with tool use behind an approval gate, sub-agents, MCP servers, plugins
and skills, auto-configured against the local instance. It also ships a
shell integration script for zsh, bash and fish that binds Ctrl+Space.
That is one of the larger user-facing changes in the release and it was
missing from both the post and the release-notes highlights. Added a
section after 3D generation, including the breaking changes for anyone
who had habits around the old REPL: `/clear` is gone in favour of
`/compact`, and a model switch now keeps the conversation.
While in the intro, corrected the counts. The post said 374 pull
requests in twenty-one days, which was accurate when it was drafted on
the 4th but not once v4.8.0 was tagged on the 5th. The published release
notes say 386 in twenty-two days, and the intro now matches them rather
than contradicting them.
For the record, neither figure is exactly right: `git log --format=%s
v4.7.1..v4.8.0 | grep -cE '\(#[0-9]+\)$'` counts 388 squash-merged pull
requests, and 389 from v4.7.0. The notes were cut before the last few
landed. Matching the published notes was the priority here, since that
is the artifact everyone else quotes, and 386 is the number already in
circulation.
Assisted-by: Claude:claude-opus-5 [Claude Code]
Signed-off-by: Ettore Di Giacinto <mudler@localai.io>
Co-authored-by: Ettore Di Giacinto <mudler@localai.io>
The two landing-page clips predated the v4.8.0 interface work (#11288,
#11305, #11307): the gallery clip showed the retired light-theme Install
Models table, and the hero clip toured the Nodes pages in a full browser
window while its caption promised a chat completion on CPU.
Both are re-recorded from a real local-ai built from v4.8.0, dark theme,
app chrome only:
- hero-ui.mp4: a chat completion on lfm2.5-1.2b-instruct streaming on
CPU with the live tok/s meter, so the caption now matches the footage.
The poster frame is regenerated from the new clip.
- gallery.mp4: the Discover rail and detail pane, the hardware
recommendation lanes, the VRAM-by-context chart, and a real install
with the live progress banner.
The hand-typed model count moves from 1,585 to 1,255 in the three places
it appears, matching the distinct-model count the recorded UI shows on
screen. The 3d-generation clip is untouched: the post-capture UI changes
do not show in its footage.
Signed-off-by: Ettore Di Giacinto <mudler@localai.io>
Assisted-by: Claude Code:claude-fable-5
Add the official Q4_K_M and Q8_0 GGUF builds with their matching vision projectors so the compact agentic model can be installed through LocalAI.
Assisted-by: Codex:gpt-5 [web]
Co-authored-by: localai-org-maint-bot <306269227+localai-org-maint-bot@users.noreply.github.com>
Cosign v2.4.1 does not select the Sigstore bundle format by default, while LocalAI's verifier only consumes OCI bundle referrers. Request the format explicitly for both registries and guard the producer contract with a shell regression test.
Document strict backend integrity configuration and release-tag identities for operators.
Assisted-by: Codex:gpt-5
Co-authored-by: localai-org-maint-bot <306269227+localai-org-maint-bot@users.noreply.github.com>
Register JSON API exchanges before their handlers run so the traces dashboard can surface active work. Replace the live entry with the completed persisted record under the same ID, and clean it up if a handler panics.
Assisted-by: Codex:gpt-5
Co-authored-by: localai-org-maint-bot <306269227+localai-org-maint-bot@users.noreply.github.com>
The runtime section described the small core and on-demand backends but
never stated the simple fact readers look for: one model can run on
llama.cpp while the next loads on vLLM, SGLang or MLX, behind the same
endpoint, and switching is one line in the model's config.
Signed-off-by: Ettore Di Giacinto <mudler@localai.io>
Assisted-by: Claude Code:claude-fable-5
* docs(blog): new DeepSeek and Laguna numbers, visuals, humanizer pass
vllm.cpp master moved 26 commits past what the post was written against,
and two results changed enough to matter. Both came from the same lever:
staging weights device-resident at load instead of reading them from the
GGUF mmap over unified memory, which the GB10 reads about 20% slower per
GEMV than device memory.
- DeepSeek-V4-Flash against DwarfStar: 0.997x parity becomes 1.144x
ahead, 18.69 vs 16.33 tok/s decode, same generated tokens.
- Laguna-XS-2.1 against vLLM: 87% becomes 1.03x, 44.46 vs 43.10 tok/s.
New row in the scoreboard.
Adds three visuals. A chart of throughput against every reference engine,
which is worth having now that the spread is 0.976 to 1.144 rather than a
flat line at parity. The Activity page with four installs running, and the
model detail pane with all four pocket-35b variants. Both screenshots were
recaptured on 2026-08-04 because #11288, #11305, #11307 and #11222 had all
changed those pages since the earlier set.
llama.cpp is deliberately absent from the chart: its 1.18x is a prefill
ratio, and putting it on the same axis as throughput ratios would be
comparing two different measurements.
Also carries the media the release notes embed, since a GitHub release
body needs URLs that survive publishing and drag-and-drop has no CLI.
Supersedes #11364.
Humanizer pass on the prose. The post had collected five exactness idioms
in one section (token-for-token, byte-exact twice, byte-identical,
token-identical). One is precision, five is a tic, so the 27B row keeps
its "token-for-token identical" where identical output is the actual
claim and the rest say what they mean. That also fixed a hyphen in
predicate position ("is token-identical").
Signed-off-by: Ettore Di Giacinto <mudler@localai.io>
Assisted-by: Claude Code:claude-opus-5 [Read] [Edit] [Bash]
* docs(blog): redraw the benchmark chart as a branded card
The Flint bar chart was generic: default palette, no brand, and drawn
from zero, which made five ratios between 0.976 and 1.144 look like five
bars of roughly equal length.
Redrawn in the style of recorder-for-agents' render-card.sh cards, the
same shape as the vllm.cpp README GIF. Palette taken from the two logos
rather than invented (LocalAI navy #0E2632 and teal #469AAF, vllm.cpp
teal #3AB4CA), SVG generated by a small JS loop so the geometry is exact
at any scale, headless Chrome to PNG at 2x.
The substantive change is that bars now run from the 1.00 parity line
instead of from zero. Deviation is what the data is about, so DeepSeek's
+14.4% and MLX-LM's -2.4% are both legible, and the one row that is
behind is the one row in amber. Each bar carries its ratio and the raw
measurement under it.
Keeps the .html source next to the .png so the chart is editable later:
change a number, re-run render-card.sh.
Signed-off-by: Ettore Di Giacinto <mudler@localai.io>
Assisted-by: Claude Code:claude-opus-5 [Read] [Edit] [Bash]
---------
Signed-off-by: Ettore Di Giacinto <mudler@localai.io>
Co-authored-by: Ettore Di Giacinto <mudler@localai.io>
This is the text users read in the backends list and the gallery, and it
was the last place still describing vllm.cpp as "a from-scratch C++20
port of vLLM created and maintained by the LocalAI team" with no
indication of maturity.
Three corrections, matching the v4.8 release notes and blog post:
- It leads with ALPHA. These are alpha development builds and llama-cpp
stays the recommendation for production, which is the single most
useful thing to know before clicking install.
- It is maintained by the LocalAI team but developed in its own
repository and usable without LocalAI. vLLM is named for what it
actually is, the reference implementation that output is checked
against and benchmarked against, rather than just the thing that was
ported.
- It records the featureset that has grown past vLLM: GGUF loading,
speculative decoding and KV offload, alongside the architecture and
hardware coverage that were already listed.
Also notes that the project is expected to be renamed, with the new name
still to be decided, so anyone who installs it now is not surprised
later.
vllm-cpp-development inherits all of this through the YAML anchor, so
both entries are covered by the one edit. Verified the file still parses
and that both entries carry the new text.
Assisted-by: Claude Code:claude-opus-5 [Read] [Edit] [Bash]
Signed-off-by: Ettore Di Giacinto <mudler@localai.io>
Co-authored-by: Ettore Di Giacinto <mudler@localai.io>
* docs(blog): final figures for the 4.8 post, and the MLX provider
The cycle closed at 374 PRs over twenty-one days, not the 321 over
eighteen the post was written against. Corrects the summary, the opening
line, the contributor count and the gallery total, and moves the date to
the day the release is cut.
Adds the MLX GEMM provider (#11137), which merged after the post was
written and is the one number an Apple Silicon reader wants: 1.54x to
2.19x on an M4 with time to first token roughly halving, both arms
toggled on one binary. The +/-10% caveat travels with the table rather
than being left in the PR.
Two lines edited against the no-ai-slop skill while I was in the file,
the same pass #11324 ran over the engines post:
- The opener balanced two clauses across a colon and closed on "without
lying to you", which is the built-to-be-quoted shape readers picked
out of the HN thread. It is a flat statement now.
- "This is a new modality rather than a new backend under an existing
one" is a binary contrast that says nothing the next clause does not.
Signed-off-by: Ettore Di Giacinto <mudler@localai.io>
Assisted-by: Claude Code:claude-opus-5 [Read] [Edit] [Bash]
* docs(blog): call vllm.cpp alpha, and finish the no-ai-slop pass
vllm.cpp is not a released backend and the post read like it was. The
old wording buried the caveat in a block quote at the end of the section
and still said "first release of a young engine". It now says plainly,
before the caveat can be skipped, that these are alpha development
builds, that shipping them in 4.8 is about letting people try the thing
rather than recommending it, and that llama-cpp stays the default.
Also completes the no-ai-slop pass I had only half run. Counting the
lines built to be quoted, headings and section endings included, the post
is in reasonable shape: long flat informational stretches, tables
followed by a plain finding, headings that are labels rather than
epigram-verdicts. Three patterns survived, each one an item in eval.md:
- "and inverts that:" set the usual shape against ours across a colon.
The sentence works without the frame.
- "Two things were conflated there: a signal, which needs one line, and
the detail, which needs somewhere to put it" is a role-assignment pair.
Says what happens instead.
- "The maturity statement from the release notes is worth repeating in
full" is throat-clearing in front of a quote, and the quote is gone.
Left the rest alone. Minimum effective edit, not a rewrite.
Signed-off-by: Ettore Di Giacinto <mudler@localai.io>
Assisted-by: Claude Code:claude-opus-5 [Read] [Edit] [Bash]
* docs(blog): present vllm.cpp as a community project, with its own numbers
The post described vllm.cpp as "a from-scratch port of vLLM, written and
maintained by the LocalAI team". Two things wrong with that. It is a
community project, and it has stopped being only a port: it loads GGUF,
runs on CPU, Metal and Vulkan, ships speculative decoding and KV offload,
and its benchmark page measures against llama.cpp, MLX-LM and DwarfStar
as well as vLLM, because those are the engines it competes with on that
hardware.
vLLM's role is now stated for what it is, the reference implementation.
Correctness is checked against it and the scoreboard is kept against it.
Also flags that the name will probably change, since it is drifting far
enough that vllm.cpp will eventually mislead.
Adds real numbers from the project's own docs/BENCHMARKS.md rather than
adjectives: 1.045x vLLM at concurrency 1 on Qwen3.6-27B NVFP4 with
token-for-token identical output, 1.010x and 1.013x at c16 and c32 on the
35B MoE and behind below that, prefill 1.18x over llama.cpp on CPU
aarch64, 97.6% of MLX-LM warm total on an M4. Upstream's own caution
travels with them: it treats c2 through c32 as ties because its noise
band is 0.5% and those margins are 0.7% to 1.7%.
Every figure was checked against ~/_git/vllm.cpp/docs/BENCHMARKS.md
rather than restated from memory. The heading is marked alpha to match
the section body.
Signed-off-by: Ettore Di Giacinto <mudler@localai.io>
Assisted-by: Claude Code:claude-opus-5 [Read] [Edit] [Bash]
* docs(blog): say who maintains vllm.cpp, and add the DeepSeek Flash result
Two corrections to the previous commit.
"A community project" says nothing and was not quite true either. The
LocalAI team maintains vllm.cpp. Community-first is the intent, not a
description, so it now says that and says what backs it: its own
repository, its own docs, benchmark record and issue tracker, and it runs
without LocalAI anywhere in the picture.
Adds the DeepSeek-V4-Flash result, which makes the divergence point
better than any of the prose around it. That model does not run on vLLM
on a single GB10: every vLLM-loadable checkpoint is 156 GB or more
against a 119 GiB unified pool, and the only quant that fits is an
extreme-low-bit GGUF that vLLM cannot load. vllm.cpp reads GGUF and runs
it at 16.28 tok/s against ds4's 16.33, a parity result. Also notes MTP
speculative decoding, token-identical to vLLM's and about 4% faster at
concurrency 1.
Both figures checked against ~/_git/vllm.cpp/docs/BENCHMARKS.md.
Signed-off-by: Ettore Di Giacinto <mudler@localai.io>
Assisted-by: Claude Code:claude-opus-5 [Read] [Edit] [Bash]
* docs(blog): lead the DeepSeek result with what we run, not with what vLLM cannot
The previous version opened on "that model does not run on vLLM on a
single GB10 at all". Wrong emphasis twice over: it makes a strong
negative claim about another project the headline, and it buries the
actual result, which is that vllm.cpp runs DeepSeek-V4-Flash at roughly
2-bit (IQ2_XXS mixed, about 80 GB) on a single DGX Spark and decodes at
16.28 tok/s against DwarfStar's 16.33.
The size constraint is still there, stated as the reason the quant is
what it is rather than as a point about vLLM: at 300B+ total parameters
even a 4-bit checkpoint is 156 GB or more, so a 2-bit GGUF is what fits
the Spark's 119 GiB unified pool.
The table row now names the quant and the box (IQ2_XXS, one DGX Spark)
instead of just "GGUF, GB10", since that is the part a reader with a
Spark wants.
Figures unchanged and still from ~/_git/vllm.cpp/docs/BENCHMARKS.md.
Signed-off-by: Ettore Di Giacinto <mudler@localai.io>
Assisted-by: Claude Code:claude-opus-5 [Read] [Edit] [Bash]
* docs(blog): say the new name is undecided
"The name will probably change at some point" invited the obvious
question. It now says the rename is expected and the name is still to be
decided, which is the actual state.
Signed-off-by: Ettore Di Giacinto <mudler@localai.io>
Assisted-by: Claude Code:claude-opus-5 [Read] [Edit] [Bash]
---------
Signed-off-by: Ettore Di Giacinto <mudler@localai.io>
Co-authored-by: Ettore Di Giacinto <mudler@localai.io>
Replace copied HauhauCS base-model text with metadata for the actual Genesis Hermes V6 artifact and link its upstream base model.
Assisted-by: Codex:gpt-5 [Hugging Face]
Co-authored-by: localai-org-maint-bot <306269227+localai-org-maint-bot@users.noreply.github.com>
No -gpu-rocm-hipblas-llama-cpp image has been published since 2026-08-01.
Every build since has been killed by GitHub at exactly its 6h job limit:
job 91830652349 cancelled 6h00m (2026-08-04)
job 91763226161 cancelled 6h00m (2026-08-03)
job 91466626154 cancelled 6h00m (2026-08-02 full matrix)
The registry shows the damage: master-gpu-rocm-hipblas-llama-cpp last
built 2026-08-01 05:53, latest-gpu-rocm-hipblas-llama-cpp 2026-07-15,
against master-cpu-llama-cpp which is current.
Same cause as #11321, different mechanism. Since #11255 every x86 GPU
image also builds ggml's CPU_ALL_VARIANTS matrix. SYCL died because icpx
stalls on one translation unit; ROCm dies on volume. hipcc compiles the
HIP kernels once per entry in AMDGPU_TARGETS, and that list is eleven
architectures (gfx908, gfx90a, gfx942, gfx950, gfx1030, gfx1100, gfx1101,
gfx1102, gfx1151, gfx1200, gfx1201). The CPU matrix lands on top of that.
The numbers are unambiguous. The same job took 2h27m in the 2026-07-26
full matrix, before #11255. #11255 merged 2026-08-01 07:26, an hour and a
half after the last image was published, and it has been 6h00m ever since.
The tail of the last run shows it 61% through ggml-hip at the 83 minute
mark, still building HIP template instances.
Route hipblas to the portable fallback, exactly as #11321 did for SYCL and
for the same practical reason: it is what these images shipped before
#11255, and run.sh already prefers *-cpu-all when present and falls back
otherwise. Expected to restore the 2h27m build with room to spare.
Not fixed here: the CPU variant matrix is genuinely wanted on ROCm for
partial offload. Getting it needs the build to fit in 6h, which means
trimming AMDGPU_TARGETS or splitting the job per architecture. Both are
larger changes than unbreaking the image, and neither should ride along
with a build that is currently not shipping at all.
Verified: make test-build-scripts passes, including the extended
llama-cpp-build-target_test.sh. bonsai is unaffected (own compile script,
ROCm builds in 1h52m) and turboquant has no hipblas variant.
Assisted-by: Claude Code:claude-opus-5 [Read] [Edit] [Bash]
Signed-off-by: Ettore Di Giacinto <mudler@localai.io>
Co-authored-by: Ettore Di Giacinto <mudler@localai.io>
* feat(vllm-cpp): enable and vendor the MLX GEMM provider on darwin/metal
The darwin vllm-cpp image built the Metal backend with vllm.cpp's native MSL
GEMM only. vllm.cpp also ships an optional MLX provider for the dense GEMM,
kept OFF upstream because it costs a ~19 MB libmlx.dylib plus a ~105 MB
mlx.metallib, on the stated position that it must earn that cost by
measurement.
Measured on an Apple M4 (16 GiB, macOS 26.5.2) it does. One binary, arms
toggled with VT_OP_PROVIDER_DISABLE=mlx so there is no build-difference
confound, Qwen3-1.7B-bf16 p=512 g=128, 2 reps, arm order alternated per rep:
B=1 5.79 vs 3.08 agg tok/s (1.88x) TTFT 3.32 s vs 7.68 s
B=8 25.70 vs 13.69 (1.88x) TTFT 13.95 s vs 34.38 s
B=16 38.65 vs 17.69 (2.19x) TTFT 18.33 s vs 54.48 s
Peak RSS is unchanged (6.65 to 7.50 GB in both arms) and the output is
bit-identical: vllm.cpp's three-way parity test measures mlx-vs-msl NMSE of 0
on all six shapes, and mlx-vs-cpu equal to msl-vs-cpu, against a 5e-4 bar. MLX
serves the dense GEMM alone; paged attention stays vllm.cpp's own kernel
because MLX has no paged-KV primitive. Full disposition, including the
INDICATIVE status and the isolation actually achieved, is in vllm.cpp
docs/BENCHMARKS.md "MLX GEMM provider A/B on Apple M4".
Build: MLX comes from the pinned prebuilt pip wheel (MLX_VERSION, default
0.29.3) into a venv under the backend dir. Building MLX from source needs
`xcrun metal`, i.e. a full Xcode the macOS runners do not have, while the wheel
ships include/, lib/libmlx.dylib and the compiled metallib ready to link. The
install is a stamp FILE rather than a phony target, because a phony
prerequisite is always newer than libvllm and would re-link it every
invocation. VLLM_CPP_MLX=off restores the previous Metal build.
Packaging vendors libmlx.dylib, mlx.metallib and MLX's MIT license into
package/lib/. Three things this had to get right, each verified on the M4
before it was written rather than after:
1. libvllm.dylib links @rpath/libmlx.dylib and its build-time LC_RPATH points
inside the build venv, a path no user has. Every build rpath is deleted
and replaced with @loader_path/lib.
2. MLX loads its metallib from beside its OWN dylib, so both files must land
in the same directory or every Metal op fails with "Failed to load the
default metallib".
3. install_name_tool invalidates the code signature and macOS refuses to load
an arm64 image with a stale one, so the patched library is re-signed
ad-hoc.
Verified end to end on the M4 by building through this Makefile and running the
packaged artifact: `DYLD_PRINT_LIBRARIES` resolves libmlx from package/lib/,
`codesign -v` passes, no build-venv path survives in the load commands, and a
real generation runs with the provider selected (op=65 selected=mlx) and zero
metallib failures. A missing rpath now fails the build instead of the user's
first inference.
Cost: the darwin vllm-cpp image grows by about 124 MB.
Signed-off-by: Ettore Di Giacinto <mudler@localai.io>
Assisted-by: Claude Code:claude-opus-5 [ClaudeCode]
* fix(vllm-cpp): default the MLX GEMM provider OFF on darwin
This branch opened with VLLM_CPP_MLX=on, justified by an A/B that measured the
MLX provider at 1.88x to 2.19x against the native MSL GEMM. That measurement was
correct when taken and is now stale: vllm.cpp's own Metal kernels have improved
several-fold since, through mma prefill attention, a vectorised decode V
accumulation, vectorised attention staging, a fused qk-norm-RoPE preamble and a
simdgroup-per-row softmax. The native path MLX was compared against no longer
exists.
Re-measured on the same Apple M4, in the same binary, with the arms toggled by
VT_OP_PROVIDER_DISABLE=mlx, on Qwen3-1.7B-bf16 warm at p=512 g=128:
MLX provider ON prefill TTFT 1370 ms warm throughput 11.98 tok/s
MLX provider OFF prefill TTFT 1400 ms warm throughput 22.06 tok/s
Shipping the previous default would have halved Apple Silicon throughput.
MLX's steel GEMM is still about 20% faster than ours in isolation, but the
provider pays a per-op mx::eval synchronisation plus an output memcpy, because it
cannot write into our buffer. Across prefill's roughly 112 GEMMs that overhead
leaves a 2% gain; on decode, where the same synchronisation is paid once per
matmul per token, it costs 46%. The option is kept for prefill-dominated
workloads, where the margin is small but real.
The README section is rewritten rather than patched: it previously presented the
stale table as the reason for the default, so leaving it in place would have made
the new default look arbitrary.
Assisted-by: Claude Code:claude-opus-5 [ClaudeCode]
Signed-off-by: Ettore Di Giacinto <mudler@localai.io>
* feat(vllm-cpp): bump vllm.cpp and default MLX ON, gated to prefill
Bumps VLLM_CPP_VERSION from 9e1c9025 to eec09bed and turns VLLM_CPP_MLX back on.
These two must move together, which is why they are one commit.
Upstream now shape-gates the MLX provider to prefill: it declines m < 2, which is
exactly the decode GEMV. MLX's steel GEMM wins prefill, 524.5 ms of TTFT against
602 for the native path, but loses decode badly because the provider pays an
mx::eval synchronisation and an output memcpy on every call while decode makes
about 112 calls per token. Ungated it does both; gated it does only the good half.
Measured on an Apple M4 with Qwen3-1.7B-bf16 warm at p=512 g=128:
MLX gated to prefill (pin >= 89c46aeb) TTFT 524.5 ms 24.40 tok/s, 99.1% of MLX-LM
MLX ungated (older pins) TTFT 537 ms 12.7 tok/s
MLX off TTFT 602 ms 23.9 tok/s
This branch briefly defaulted the provider off, which was the correct call for an
ungated provider at the old pin. The gate is what makes on correct again, so the
pin and the flag are coupled: rolling VLLM_CPP_VERSION back before 89c46aeb while
leaving MLX on would select the middle row and roughly halve throughput. Both the
Makefile comment and the README state that dependency explicitly.
The bump also brings six Metal kernels landed upstream since the old pin — mma
prefill attention, a vectorised decode V accumulation, vectorised attention
staging, a fused qk-norm-RoPE preamble, a simdgroup-per-row softmax and a
simdgroup-per-head preamble — which take the non-MLX Metal path from 89.4% to
96.4% of MLX-LM on their own.
One caveat, recorded in the README: MLX's GEMM is not bit-identical to the native
kernel, so an MLX build produces a different greedy sequence than a non-MLX build.
That is a property of the provider rather than of the gate and predates this
packaging.
Assisted-by: Claude Code:claude-opus-5 [ClaudeCode]
Signed-off-by: Ettore Di Giacinto <mudler@localai.io>
* docs(vllm-cpp): correct the MLX-gated figure to 97.6%, from 99.1%
The previous commit quoted 99.1% of MLX-LM for the prefill-gated MLX build. That
figure divided by a two-run MLX-LM baseline, 27.135 and 27.744 generation tok/s
averaged to 27.44. Re-measured interleaved with ours over four ABBA blocks,
MLX-LM's decode is 27.848 with a 0.34% spread across six runs, so the 27.135 was
an outlier and averaging it in overstated us by roughly 1.5 points.
Corrected: the gated configuration is 24.37 tok/s, or 97.6% of MLX-LM, and the
MLX-off build is 23.9 tok/s or 95.9%. Prefill TTFT is unchanged at 524.5 ms
against MLX-LM's 532.6, so we remain about 1.5% faster there.
Nothing else changes. MLX still wins prefill and loses decode, the shape gate is
still the right disposition, and the pin and the flag are still coupled. The gate
is worth about 1.7 points over the MLX-off build rather than 2.7.
Assisted-by: Claude Code:claude-opus-5 [ClaudeCode]
Signed-off-by: Ettore Di Giacinto <mudler@localai.io>
* fix(vllm-cpp): pin MLX gate from mainline
The previous pin was a merge commit from the experimental C ABI v9 branch. Pin the same MLX prefill gate on upstream main so the backend build does not pull unrelated ABI v9 work into every platform variant.
Assisted-by: Codex:gpt-5 [systematic-debugging]
* fix(vllm-cpp): restore backend build portability
Keep the current master pin when enabling MLX so every backend variant builds against the known-good vllm.cpp revision. Suppress Apple clang’s GNU constant-folding diagnostic for Objective-C++ Metal compilation only, since upstream treats warnings as errors.
Assisted-by: Codex:gpt-5 [systematic-debugging]
* fix(vllm-cpp): demote MLX header VLA warning
MLX 0.29.3 headers trigger Apple clang's gnu-folding-constant diagnostic in the Objective-C++ provider. Keep the diagnostic visible while exempting only it from vllm.cpp's global warnings-as-errors policy.
Assisted-by: Codex:gpt-5 [systematic-debugging]
* fix(vllm-cpp): suppress MLX header VLA warning
Target-level Objective-C++ -Werror is appended after the directory flags, so a no-error demotion is re-promoted. Disable this single warning for the MLX header while keeping every other warning fatal.
Assisted-by: Codex:gpt-5 [systematic-debugging]
* fix(vllm-cpp): pin source-scoped MLX warning fix
Move the AppleClang warning exception into vllm.cpp where its target warning policy is defined, and pin LocalAI to that source-scoped fix.
Assisted-by: Codex:gpt-5
* fix(vllm-cpp): pin effective MLX warning suppression
The source-scoped no-error flag was overridden by the target warning policy. Pin the companion vllm.cpp change that disables only the MLX header diagnostic for its Objective-C++ translation unit.
Assisted-by: Codex:gpt-5
* fix(vllm-cpp): pin diagnostic pragma fix
Pin the companion vllm.cpp correction that scopes the AppleClang folding warning suppression inside the MLX translation unit, after command-line warning policy.
Assisted-by: Codex:gpt-5 [systematic-debugging]
* fix(vllm-cpp): pin remaining Darwin build fixes
Advance the MLX-enabled backend to the vllm.cpp revision already validated by the dependency update branch. This includes the feature guards and AppleClang pragma boundary needed by the Darwin build.
Assisted-by: Codex:gpt-5 [systematic-debugging]
* fix(vllm-cpp): pin MLX system dependency boundary
Pin the companion vllm.cpp change that models MLX as an imported system dependency, keeping third-party header diagnostics out of the project's warnings-as-errors policy while retaining fatal warnings for project sources.
Assisted-by: Codex:gpt-5 [Codex]
* fix(vllm-cpp): pin scoped MLX warning guard
Advance vllm.cpp to the companion fix that keeps MLX headers on a SYSTEM dependency and scopes AppleClang folding-constant suppression to the external includes.
Assisted-by: Codex:gpt-5 [systematic-debugging] [test-driven-development]
* fix(vllm-cpp): use available MLX wheel
MLX 0.29.3 is no longer available to the Darwin runner, so the backend build stopped before CMake. Pin the first available compatible wheel and keep the documented default in sync.
Assisted-by: Codex:gpt-5
---------
Signed-off-by: Ettore Di Giacinto <mudler@localai.io>
Co-authored-by: Ettore Di Giacinto <mudler@localai.io>
Co-authored-by: localai-org-maint-bot <306269227+localai-org-maint-bot@users.noreply.github.com>
* ⬆️ Update antirez/ds4
Signed-off-by: github-actions[bot] <41898282+github-actions[bot]@users.noreply.github.com>
* fix(ds4): link upstream CUDA MMQ objects
The updated ds4 CUDA object now calls into the vendored MMQ implementation. Build and link those objects into both the gRPC server and distributed worker.
Assisted-by: Codex:gpt-5 [Codex]
---------
Signed-off-by: github-actions[bot] <41898282+github-actions[bot]@users.noreply.github.com>
Co-authored-by: mudler <2420543+mudler@users.noreply.github.com>
Co-authored-by: localai-org-maint-bot <306269227+localai-org-maint-bot@users.noreply.github.com>
Wait for observable loop events instead of budgeting hundreds of milliseconds for scheduler timing. Keep a short bounded overlap observation for the two-leader exclusion check.
Assisted-by: Codex:gpt-5
Co-authored-by: localai-org-maint-bot <306269227+localai-org-maint-bot@users.noreply.github.com>
