Stored Completions

Set "store": true on a chat-completions request to persist the finished completion, then read it back later by its chatcmpl-... id. Storage follows OpenAI’s store semantics and happens after the response has been fully delivered — it never adds latency to the request, and a storage failure never fails the completion. Streamed completions are stored too: the final chat.completion object is assembled from the streamed chunks (content, tool calls, finish_reason, usage). Stored completions are org-scoped: they are only visible to (and deletable by) API keys of the organization that created them.

Store a completion

curl -X POST https://api.stage.peachdeskai.com/api/v1/voice-engine/llm/v1/chat/completions \
  -H "X-API-Key: bibha_your_api_key" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "brahma-30b",
    "messages": [
      {"role": "user", "content": "Summarize PeachDesk in one sentence."}
    ],
    "max_completion_tokens": 256,
    "store": true,
    "metadata": {"session_id": "sess_412", "channel": "web"}
  }'

The response is a normal completion; note its id — that is the retrieval key.

`metadata` limits

metadata is an optional object of string keys to string values stored alongside the completion:

Limit	Value
Maximum pairs	16
Key length	≤ 64 characters
Value type / length	string, ≤ 512 characters

Violations are rejected with a 422 OpenAI error envelope (param: "metadata") before the request is run — nothing is generated and no tokens are consumed.

Retrieve a stored completion

GET /api/v1/voice-engine/llm/v1/chat/completions/{completion_id}

curl https://api.stage.peachdeskai.com/api/v1/voice-engine/llm/v1/chat/completions/chatcmpl-a1b2c3 \
  -H "X-API-Key: bibha_your_api_key"

Returns the stored chat.completion object verbatim — the same shape the original request returned (for streamed requests, the assembled final object including usage). An unknown id, an id stored by another organization, or a completion made without store: true all answer the same OpenAI 404 envelope:

{
  "error": {
    "message": "Completion chatcmpl-a1b2c3 not found",
    "type": "invalid_request_error",
    "param": null,
    "code": "not_found"
  }
}

Delete a stored completion

DELETE /api/v1/voice-engine/llm/v1/chat/completions/{completion_id}

curl -X DELETE https://api.stage.peachdeskai.com/api/v1/voice-engine/llm/v1/chat/completions/chatcmpl-a1b2c3 \
  -H "X-API-Key: bibha_your_api_key"

{
  "object": "chat.completion.deleted",
  "id": "chatcmpl-a1b2c3",
  "deleted": true
}

Deleting an id that does not exist (or belongs to another organization) returns the same 404 envelope as retrieval.

Unsupported endpoints

Only retrieve and delete are implemented. The rest of OpenAI’s stored-completions surface is not supported:

OpenAI endpoint	Status
`GET /v1/chat/completions` (list)	Not supported
`POST /v1/chat/completions/{id}` (update metadata)	Not supported
`GET /v1/chat/completions/{id}/messages`	Not supported

Track completion ids on your side (for example in your own database, keyed by the metadata you attach) rather than relying on a list endpoint.

Resources

Authentication & Errors

Voice Engine API

Bibha BRAHMA LLM

Bibha VAANI TTS

Bibha LIPI STT

Store a completion

`metadata` limits

Retrieve a stored completion

Delete a stored completion

Unsupported endpoints

​Store a completion

​metadata limits

​Retrieve a stored completion

​Delete a stored completion

​Unsupported endpoints

Store a completion

`metadata` limits

Retrieve a stored completion

Delete a stored completion

Unsupported endpoints