feat(model-config): completion_type array + CRUD endpoints

backend/app/alembic/versions/064_completion_type_array.py

@@ -0,0 +1,49 @@
+"""completion_type scalar to array
+
+Revision ID: 064
+Revises: 063
+Create Date: 2026-05-26 00:00:00.000000
+
+"""
+
+from alembic import op
+
+revision = "064"
+down_revision = "063"
+branch_labels = None
+depends_on = None
+
+
+def upgrade():
+    # Drop scalar index before altering column
+    op.execute("DROP INDEX IF EXISTS global.ix_model_config_provider_type_active")
+
+    # Convert scalar enum to array, wrapping existing values
+    op.execute(
+        """
+        ALTER TABLE global.model_config
+            ALTER COLUMN completion_type TYPE global.completion_type_enum[]
+                USING ARRAY[completion_type]
+        """
+    )
+
+    # GIN index for efficient array containment queries
+    op.execute(
+        "CREATE INDEX ix_model_config_completion_type ON global.model_config USING gin (completion_type)"
+    )
+
+
+def downgrade():
+    op.execute("DROP INDEX IF EXISTS global.ix_model_config_completion_type")
+
+    op.execute(
+        """
+        ALTER TABLE global.model_config
+            ALTER COLUMN completion_type TYPE global.completion_type_enum
+                USING completion_type[1]
+        """
+    )
+
+    op.execute(
+        "CREATE INDEX ix_model_config_provider_type_active ON global.model_config (provider, completion_type, is_active)"
+    )

backend/app/api/docs/model_config/create_models.md

@@ -0,0 +1,53 @@
+Create one or more model configurations.
+
+Accepts a single object or an array. Response is always an array.
+
+**Required:** `provider`, `model_name`, `completion_type`, `config`
+**Optional:** `input_modalities`, `output_modalities`, `pricing`, `is_active`
+
+`(provider, model_name)` must be unique.
+
+### Example (single)
+
+```json
+{
+  "provider": "google",
+  "model_name": "gemini-2.5-flash",
+  "completion_type": ["text", "stt"],
+  "config": { "temperature": { "type": "float", "default": 1.0, "min": 0.0, "max": 2.0 } },
+  "input_modalities": ["TEXT", "AUDIO"],
+  "output_modalities": ["TEXT"],
+  "pricing": {
+    "response": { "input_token_cost": 0.3, "output_token_cost": 2.5 }
+  },
+  "is_active": true
+}
+```
+
+### Example (multiple)
+
+```json
+[
+  {
+    "provider": "sarvamai",
+    "model_name": "saaras:v3",
+    "completion_type": ["stt"],
+    "config": {},
+    "input_modalities": ["AUDIO"],
+    "output_modalities": ["TEXT"]
+  },
+  {
+    "provider": "elevenlabs",
+    "model_name": "scribe_v2",
+    "completion_type": ["stt"],
+    "config": {},
+    "input_modalities": ["AUDIO"],
+    "output_modalities": ["TEXT"]
+  }
+]
+```
+
+### Errors
+
+- `422` — Validation error.
+- DB integrity error on duplicate `(provider, model_name)`.

backend/app/api/docs/model_config/delete_model.md

@@ -0,0 +1,7 @@
+Permanently delete a model configuration. Hard delete — cannot be undone.
+
+To hide a model instead, use `PATCH` with `is_active: false`.
+
+### Errors
+
+- `404` — Model not found.

backend/app/api/docs/model_config/get_model.md

@@ -1,67 +1,5 @@
-## Endpoint
+Get a single model configuration by `provider` and `model_name`.
 
-**GET** `/api/v1/models/{provider}/{model_name}`
+### Errors
 
-Retrieve a specific model configuration by provider and model name.
-
-Returns model details including supported config parameters, input/output modalities, pricing, and active status.
-
-### Path Parameters
-
-- **`provider`** (required) — Provider name (e.g. `openai`, `google`)
-- **`model_name`** (required) — Model name (e.g. `gpt-4o`, `gpt-4o-mini`)
-
-### Example Response
-
-```json
-{
-  "success": true,
-  "data": {
-    "id": 2,
-    "provider": "openai",
-    "model_name": "gpt-4o",
-    "config": {
-      "temperature": {
-        "type": "float",
-        "default": 1.0,
-        "min": 0.0,
-        "max": 2.0,
-        "description": "Controls randomness. Lower = more deterministic."
-      },
-      "top_p": {
-        "type": "float",
-        "default": 1.0,
-        "min": 0.0,
-        "max": 1.0,
-        "description": "Nucleus sampling. Use either this or temperature, not both."
-      },
-      "max_output_tokens": {
-        "type": "int",
-        "default": 2048,
-        "min": 1,
-        "max": 32768,
-        "description": "Max tokens in the response."
-      }
-    },
-    "input_modalities": ["TEXT", "IMAGE"],
-    "output_modalities": ["TEXT"],
-    "pricing": {
-      "response": {
-        "input_token_cost": 2.5,
-        "output_token_cost": 10
-      },
-      "batch": {
-        "input_token_cost": 1.25,
-        "output_token_cost": 5
-      }
-    },
-    "is_active": true,
-    "inserted_at": "2026-03-12T00:00:00",
-    "updated_at": "2026-03-12T00:00:00"
-  }
-}
-```
-
-### Error Response
-
-- `404 Not Found` — Model not found for the given `provider` and `model_name`.
+- `404` — Model not found.

backend/app/api/docs/model_config/list_models.md

@@ -1,74 +1,3 @@
-## Endpoint
+List active model configurations.
 
-**GET** `/api/v1/models`
-
-Retrieve a list of all active model configurations.
-
-Returns model details including provider, model name, supported config parameters, input/output modalities, pricing, and active status.
-
-Optionally filter by provider (e.g. openai, google).
-
-### Query Parameters
-
-- **`provider`** (optional) — Filter by provider name (e.g. `openai`, `google`)
-- **`skip`** (optional, default 0) — Number of records to skip for pagination
-- **`limit`** (optional, default 100, max 100) — Maximum number of records to return
-
-### Example Response
-
-```json
-{
-  "success": true,
-  "metadata": {
-    "has_more": true
-  },
-  "data": {
-    "data": [
-      {
-        "id": 1,
-        "provider": "openai",
-        "model_name": "gpt-4o-mini",
-        "config": {
-          "temperature": {
-            "type": "float",
-            "default": 1.0,
-            "min": 0.0,
-            "max": 2.0,
-            "description": "Controls randomness. Lower = more deterministic."
-          },
-          "top_p": {
-            "type": "float",
-            "default": 1.0,
-            "min": 0.0,
-            "max": 1.0,
-            "description": "Nucleus sampling. Use either this or temperature, not both."
-          },
-          "max_output_tokens": {
-            "type": "int",
-            "default": 2048,
-            "min": 1,
-            "max": 32768,
-            "description": "Max tokens in the response."
-          }
-        },
-        "input_modalities": ["TEXT", "IMAGE"],
-        "output_modalities": ["TEXT"],
-        "pricing": {
-          "response": {
-            "input_token_cost": 0.15,
-            "output_token_cost": 0.6
-          },
-          "batch": {
-            "input_token_cost": 0.075,
-            "output_token_cost": 0.3
-          }
-        },
-        "is_active": true,
-        "inserted_at": "2026-03-12T00:00:00",
-        "updated_at": "2026-03-12T00:00:00"
-      }
-    ],
-    "count": 1
-  }
-}
-```
+Filter by `provider` (`openai` | `google` | `sarvamai` | `elevenlabs`). Paginate with `skip` / `limit` (max 100). Only active models returned, sorted by `provider`, `model_name`. `metadata.has_more` flags more records.

backend/app/api/docs/model_config/list_models_grouped.md

@@ -1,52 +1,5 @@
-## Endpoint
+List active models grouped by provider.
 
-**GET** `/api/v1/models/grouped`
+Returns a dict keyed by provider; each value is the list of that provider's active models.
 
-Retrieve active models grouped by provider.
-
-Supports pagination of model rows before grouping:
-- `skip` (default `0`)
-- `limit` (default `100`, max `100`)
-
-Returns a dictionary where each key is a provider present in the paginated slice, and each value is a list of active model configurations for that provider.
-Includes `metadata.has_more` when additional model rows exist.
-
-### Example Response
-
-```json
-{
-  "success": true,
-  "metadata": {
-    "has_more": true
-  },
-  "data": {
-    "openai": [
-      {
-        "id": 2,
-        "provider": "openai",
-        "model_name": "gpt-4o",
-        "config": {
-          "temperature": {
-            "type": "float",
-            "default": 1.0,
-            "min": 0.0,
-            "max": 2.0,
-            "description": "Controls randomness. Lower = more deterministic."
-          }
-        },
-        "input_modalities": ["TEXT", "IMAGE"],
-        "output_modalities": ["TEXT"],
-        "pricing": {
-          "response": {
-            "input_token_cost": 2.5,
-            "output_token_cost": 10
-          }
-        },
-        "is_active": true,
-        "inserted_at": "2026-03-12T00:00:00",
-        "updated_at": "2026-03-12T00:00:00"
-      }
-    ]
-  }
-}
-```
+Pagination (`skip` / `limit`) is applied **before** grouping — adjust `limit` if expecting many models per provider.

backend/app/api/docs/model_config/list_providers.md

@@ -1,16 +1 @@
-## Endpoint
-
-**GET** `/api/v1/models/providers`
-
-Retrieve the list of providers that currently have active models.
-
-Returns provider names sorted in ascending order.
-
-### Example Response
-
-```json
-{
-  "success": true,
-  "data": ["google", "openai"]
-}
-```
+List provider names that have at least one active model. Sorted ascending.

backend/app/api/docs/model_config/update_model.md

@@ -0,0 +1,21 @@
+Partially update a model configuration. Only fields sent are updated; omitted fields stay unchanged.
+
+**Updatable fields:** `completion_type`, `config`, `input_modalities`, `output_modalities`, `pricing`, `is_active`
+
+Arrays and objects are **replaced** (no deep merge). `provider` and `model_name` cannot be changed here.
+
+### Example
+
+```json
+{
+  "completion_type": ["text", "stt"],
+  "pricing": {
+    "response": { "input_token_cost": 0.5, "output_token_cost": 3.0 }
+  }
+}
+```
+
+### Errors
+
+- `404` — Model not found.
+- `422` — Validation error.

backend/app/api/docs/model_config/update_models.md

@@ -0,0 +1,19 @@
+Update multiple model configurations in one request.
+
+Each item must include `provider` + `model_name` to identify the target. Other fields are optional and follow the same rules as the single PATCH endpoint (replace semantics, no deep merge).
+
+Atomic — if any target is missing, no updates are applied.
+
+### Example
+
+```json
+[
+  { "provider": "google", "model_name": "gemini-2.5-flash", "completion_type": ["text", "stt"] },
+  { "provider": "sarvamai", "model_name": "saaras:v3", "is_active": false }
+]
+```
+
+### Errors
+
+- `404` — One or more targets not found.
+- `422` — Validation error.