feat(swagger): update swagger (#11585)

Signed-off-by: github-actions[bot] <41898282+github-actions[bot]@users.noreply.github.com>
Co-authored-by: mudler <2420543+mudler@users.noreply.github.com>

swagger/docs.go

@@ -2110,7 +2110,7 @@ const docTemplate = `{
         },
         "/audio/transformations/stream": {
             "get": {
-                "description": "Streams binary PCM frames in (interleaved stereo: ch0=audio, ch1=reference) and out (mono). The first message must be a JSON ` + "`" + `session.update` + "`" + ` envelope describing model + sample format + frame size + backend params. Server emits binary PCM on the same cadence.",
+                "description": "Streams binary PCM frames in (interleaved stereo: ch0=audio, ch1=reference) and out (mono). The model must support the audio_transform use case. Any-to-any models such as liquid-audio use the OpenAI Realtime API instead. The first message must be a JSON ` + "`" + `session.update` + "`" + ` envelope describing model + sample format + frame size + backend params. Server emits binary PCM on the same cadence.",
                 "tags": [
                     "audio"
                 ],
@@ -5080,6 +5080,29 @@ const docTemplate = `{
                 }
             }
         },
+        "schema.CompressionMetadata": {
+            "type": "object",
+            "properties": {
+                "compressed_tokens": {
+                    "type": "integer"
+                },
+                "compressor": {
+                    "type": "string"
+                },
+                "dropped_turns": {
+                    "type": "integer"
+                },
+                "original_tokens": {
+                    "type": "integer"
+                },
+                "overflow_recoveries": {
+                    "type": "integer"
+                },
+                "summary_tokens": {
+                    "type": "integer"
+                }
+            }
+        },
         "schema.DepthRequest": {
             "type": "object",
             "properties": {
@@ -6992,6 +7015,14 @@ const docTemplate = `{
                 "negative_prompt_scale": {
                     "type": "number"
                 },
+                "pooling": {
+                    "description": "Pooling is a LocalAI extension for /v1/embeddings: how the backend's\nper-token vectors are reduced to a single embedding. \"\" or \"backend\"\nleaves pooling to the inference backend (the pre-existing behavior);\n\"mean\", \"last\" and \"decayed_mean\" pool Go-side from raw per-token\nvectors (the backend must run with the \"pooling:none\" option, which\nmodel configs get automatically when this is set).",
+                    "type": "string"
+                },
+                "pooling_half_life_tokens": {
+                    "description": "PoolingHalfLifeTokens is a LocalAI extension for /v1/embeddings: the\nhalf-life (in tokens) of the \"decayed_mean\" pooling scheme — a token's\nweight halves every this-many positions counting back from the end of\nthe conversation. Defaults to 256 when unset.",
+                    "type": "integer"
+                },
                 "presence_penalty": {
                     "type": "number"
                 },
@@ -7129,6 +7160,9 @@ const docTemplate = `{
                 "completion_tokens": {
                     "type": "integer"
                 },
+                "compression_meta": {
+                    "$ref": "#/definitions/schema.CompressionMetadata"
+                },
                 "input_tokens": {
                     "description": "Fields for image generation API compatibility",
                     "type": "integer"

swagger/swagger.json

@@ -2107,7 +2107,7 @@
         },
         "/audio/transformations/stream": {
             "get": {
-                "description": "Streams binary PCM frames in (interleaved stereo: ch0=audio, ch1=reference) and out (mono). The first message must be a JSON `session.update` envelope describing model + sample format + frame size + backend params. Server emits binary PCM on the same cadence.",
+                "description": "Streams binary PCM frames in (interleaved stereo: ch0=audio, ch1=reference) and out (mono). The model must support the audio_transform use case. Any-to-any models such as liquid-audio use the OpenAI Realtime API instead. The first message must be a JSON `session.update` envelope describing model + sample format + frame size + backend params. Server emits binary PCM on the same cadence.",
                 "tags": [
                     "audio"
                 ],
@@ -5077,6 +5077,29 @@
                 }
             }
         },
+        "schema.CompressionMetadata": {
+            "type": "object",
+            "properties": {
+                "compressed_tokens": {
+                    "type": "integer"
+                },
+                "compressor": {
+                    "type": "string"
+                },
+                "dropped_turns": {
+                    "type": "integer"
+                },
+                "original_tokens": {
+                    "type": "integer"
+                },
+                "overflow_recoveries": {
+                    "type": "integer"
+                },
+                "summary_tokens": {
+                    "type": "integer"
+                }
+            }
+        },
         "schema.DepthRequest": {
             "type": "object",
             "properties": {
@@ -6989,6 +7012,14 @@
                 "negative_prompt_scale": {
                     "type": "number"
                 },
+                "pooling": {
+                    "description": "Pooling is a LocalAI extension for /v1/embeddings: how the backend's\nper-token vectors are reduced to a single embedding. \"\" or \"backend\"\nleaves pooling to the inference backend (the pre-existing behavior);\n\"mean\", \"last\" and \"decayed_mean\" pool Go-side from raw per-token\nvectors (the backend must run with the \"pooling:none\" option, which\nmodel configs get automatically when this is set).",
+                    "type": "string"
+                },
+                "pooling_half_life_tokens": {
+                    "description": "PoolingHalfLifeTokens is a LocalAI extension for /v1/embeddings: the\nhalf-life (in tokens) of the \"decayed_mean\" pooling scheme — a token's\nweight halves every this-many positions counting back from the end of\nthe conversation. Defaults to 256 when unset.",
+                    "type": "integer"
+                },
                 "presence_penalty": {
                     "type": "number"
                 },
@@ -7126,6 +7157,9 @@
                 "completion_tokens": {
                     "type": "integer"
                 },
+                "compression_meta": {
+                    "$ref": "#/definitions/schema.CompressionMetadata"
+                },
                 "input_tokens": {
                     "description": "Fields for image generation API compatibility",
                     "type": "integer"

swagger/swagger.yaml

@@ -750,6 +750,21 @@ definitions:
       text:
         type: string
     type: object
+  schema.CompressionMetadata:
+    properties:
+      compressed_tokens:
+        type: integer
+      compressor:
+        type: string
+      dropped_turns:
+        type: integer
+      original_tokens:
+        type: integer
+      overflow_recoveries:
+        type: integer
+      summary_tokens:
+        type: integer
+    type: object
   schema.DepthRequest:
     properties:
       dst:
@@ -2072,6 +2087,22 @@ definitions:
         type: string
       negative_prompt_scale:
         type: number
+      pooling:
+        description: |-
+          Pooling is a LocalAI extension for /v1/embeddings: how the backend's
+          per-token vectors are reduced to a single embedding. "" or "backend"
+          leaves pooling to the inference backend (the pre-existing behavior);
+          "mean", "last" and "decayed_mean" pool Go-side from raw per-token
+          vectors (the backend must run with the "pooling:none" option, which
+          model configs get automatically when this is set).
+        type: string
+      pooling_half_life_tokens:
+        description: |-
+          PoolingHalfLifeTokens is a LocalAI extension for /v1/embeddings: the
+          half-life (in tokens) of the "decayed_mean" pooling scheme — a token's
+          weight halves every this-many positions counting back from the end of
+          the conversation. Defaults to 256 when unset.
+        type: integer
       presence_penalty:
         type: number
       prompt:
@@ -2172,6 +2203,8 @@ definitions:
     properties:
       completion_tokens:
         type: integer
+      compression_meta:
+        $ref: '#/definitions/schema.CompressionMetadata'
       input_tokens:
         description: Fields for image generation API compatibility
         type: integer
@@ -4407,9 +4440,11 @@ paths:
   /audio/transformations/stream:
     get:
       description: 'Streams binary PCM frames in (interleaved stereo: ch0=audio, ch1=reference)
-        and out (mono). The first message must be a JSON `session.update` envelope
-        describing model + sample format + frame size + backend params. Server emits
-        binary PCM on the same cadence.'
+        and out (mono). The model must support the audio_transform use case. Any-to-any
+        models such as liquid-audio use the OpenAI Realtime API instead. The first
+        message must be a JSON `session.update` envelope describing model + sample
+        format + frame size + backend params. Server emits binary PCM on the same
+        cadence.'
       responses: {}
       summary: Bidirectional realtime audio transform over WebSocket.
       tags: