Gen3 Gt Api Runbooks Generic Openai Http

Start Here

Fallback row for unlisted products that can override OpenAI base URL and send bearer-authenticated requests to /api/tenant/v1/*. Recommended preset: OpenAI-compatible chat client.

Why this matters

This runbook maps Generic OpenAI-compatible HTTP client to GT API routes operators actually publish.

Details

Compatibility: Native OpenAI-compatible · Category: SDK and HTTP

Official documentation

• https://platform.openai.com/docs/api-reference/introduction
• https://swagger.io/docs/specification/authentication/bearer-authentication/
• https://developer.mozilla.org/en-US/docs/Web/HTTP/Headers/Authorization

Configuration fields

• Base URL / api_base / OPENAI_BASE_URL: https://<tenant-host>/api/tenant
• Authorization: Bearer: gtak_...
• model (request body): published-agent-alias

GT route mapping

GT route	Verdict	Client integration
GET /v1/models	native	Client HTTP GET /v1/models or static alias configuration
POST /v1/chat/completions	native	Client HTTP POST /v1/chat/completions with JSON body
POST /v1/embeddings	native	Client HTTP POST /v1/embeddings when key includes embed scope
POST /v1/audio/transcriptions	native	Client HTTP multipart POST /v1/audio/transcriptions when transcribe scope enabled
POST /v1/audio/speech	native	Client HTTP POST /v1/audio/speech when speech scope enabled
POST /v1/images/generations	native	Client HTTP POST /v1/images/generations when image scope enabled
POST /v1/conversations/files	gt_extension	Client HTTP multipart POST /v1/conversations/files with GT headers
POST /v1/datasets/{id}/files	gt_extension	Client HTTP multipart POST /v1/datasets/{id}/files
GET /v1/files/{id}	gt_extension	Client HTTP GET /v1/files/{id} polling

Not supported in this product

• Products that only speak vendor-native APIs and cannot set a custom OpenAI-compatible URL are not direct GT API clients.

Prerequisites

• Create narrowest key preset matching actual client behavior.
• Confirm client can override both base URL and bearer auth.

Setup steps

1. Configure base URL to tenant GT API origin ending in /api/tenant.
2. Set bearer token to GT API key from key creation.
3. Validate model discovery and one chat completion.
4. Add GT conversation/upload routes only if required by the integration.

GT extensions and caveats

• Layer POST /v1/conversations, upload routes, and file polling beside standard chat when needed.
• Use this row when the product is unlisted but meets the custom OpenAI URL criteria.

Validation checklist

• Client can list models and complete chat with published alias.
• Optional GT routes work when explicitly implemented.
• Key scope matches actual routes exercised.

Plain-text export

Generic OpenAI-compatible HTTP client runbook
Native OpenAI-compatible · SDK and HTTP
Recommended key preset: OpenAI-compatible chat client
Evidence: documented compatibility (vendor docs cross-check)

Fallback row for unlisted products that can override OpenAI base URL and send bearer-authenticated requests to `/api/tenant/v1/*`.

Official documentation:
- https://platform.openai.com/docs/api-reference/introduction
- https://swagger.io/docs/specification/authentication/bearer-authentication/
- https://developer.mozilla.org/en-US/docs/Web/HTTP/Headers/Authorization

Configuration fields:
- Base URL / api_base / OPENAI_BASE_URL: https://<tenant-host>/api/tenant
- Authorization: Bearer: gtak_...
- model (request body): published-agent-alias

GT route mapping:
- GET /v1/models (native): Client HTTP GET `/v1/models` or static alias configuration
- POST /v1/chat/completions (native): Client HTTP POST `/v1/chat/completions` with JSON body
- POST /v1/embeddings (native): Client HTTP POST `/v1/embeddings` when key includes embed scope
- POST /v1/audio/transcriptions (native): Client HTTP multipart POST `/v1/audio/transcriptions` when transcribe scope enabled
- POST /v1/audio/speech (native): Client HTTP POST `/v1/audio/speech` when speech scope enabled
- POST /v1/images/generations (native): Client HTTP POST `/v1/images/generations` when image scope enabled
- POST /v1/conversations/files (gt_extension): Client HTTP multipart POST `/v1/conversations/files` with GT headers
- POST /v1/datasets/{id}/files (gt_extension): Client HTTP multipart POST `/v1/datasets/{id}/files`
- GET /v1/files/{id} (gt_extension): Client HTTP GET `/v1/files/{id}` polling

Not supported in this product:
- Products that only speak vendor-native APIs and cannot set a custom OpenAI-compatible URL are not direct GT API clients.

Prerequisites:
- Create narrowest key preset matching actual client behavior.
- Confirm client can override both base URL and bearer auth.

Setup steps:
1. Configure base URL to tenant GT API origin ending in `/api/tenant`.
2. Set bearer token to GT API key from key creation.
3. Validate model discovery and one chat completion.
4. Add GT conversation/upload routes only if required by the integration.

GT extensions and caveats:
- Layer `POST /v1/conversations`, upload routes, and file polling beside standard chat when needed.
- Use this row when the product is unlisted but meets the custom OpenAI URL criteria.

Validation checklist:
- Client can list models and complete chat with published alias.
- Optional GT routes work when explicitly implemented.
- Key scope matches actual routes exercised.