> ## Documentation Index > Fetch the complete documentation index at: https://docs.infery.ai/llms.txt > Use this file to discover all available pages before exploring further. # Openai compatibility *** title: "OpenAI SDK compatibility matrix" description: "Every gateway operation: whether the official OpenAI SDK reaches it, and what to do when it doesn't." ------------------------------------------------------------------------------------------------------------------- This page is generated from the same operation list the Infery TypeScript SDK is built from, so it can't drift from what actually ships the way a hand-written table would. For the migration story and the reasoning behind the gaps, read [Migrating from OpenAI](/guides/openai-sdk-migration) first — this page is the full reference table it links to. **7 operations** work with the OpenAI SDK unchanged, **4 are partial** (they work for the common case but have a documented divergence — read the note), and **24 have no OpenAI SDK method at all** (call them over plain HTTP, or use the matching Infery SDK method named below once it ships). The **Infery SDK** column names the TypeScript SDK method that serves each operation. The `infery` package is not published to npm yet — there is deliberately no install snippet on this page. Until it ships, every "Infery SDK" and every "no"/"partial" row is reached the same way: plain HTTP. ### Chat & embeddings | Operation | Infery SDK | OpenAI SDK | Notes / how to reach it otherwise | | --------------------------- | ------------------------- | ---------- | --------------------------------- | | `POST /v1/chat/completions` | `chat.completions.create` | ✅ Yes | — | | `POST /v1/embeddings` | `embeddings.create` | ✅ Yes | — | ### Images | Operation | Infery SDK | OpenAI SDK | Notes / how to reach it otherwise | | ----------------------------- | ----------------- | ---------- | ------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- | | `POST /v1/images/generations` | `images.generate` | ⚠️ Partial | Works for the common case. A fal/replicate-routed model can instead answer 504 with a `job_id` (the request is still running and still billed) — poll `GET /v1/images/jobs/{job_id}`, which the OpenAI SDK will not do for you. | | `POST /v1/images/edits` | `images.edit` | ❌ No | The OpenAI SDK posts multipart; this endpoint takes JSON with image\_base64. | | `POST /v1/images/upscale` | `images.upscale` | ❌ No | Not reachable via the OpenAI SDK — call it over plain HTTP. | | `GET /v1/images/jobs/{id}` | `jobs.retrieve` | ❌ No | Not reachable via the OpenAI SDK — call it over plain HTTP. | ### Video | Operation | Infery SDK | OpenAI SDK | Notes / how to reach it otherwise | | ------------------------------------ | ----------------- | ---------- | ----------------------------------------------------------- | | `POST /v1/videos/generations` | `videos.submit` | ❌ No | Not reachable via the OpenAI SDK — call it over plain HTTP. | | `GET /v1/videos/generations/{jobId}` | `videos.retrieve` | ❌ No | Not reachable via the OpenAI SDK — call it over plain HTTP. | | `POST /v1/video/upscale` | `videos.upscale` | ❌ No | Not reachable via the OpenAI SDK — call it over plain HTTP. | ### Music | Operation | Infery SDK | OpenAI SDK | Notes / how to reach it otherwise | | ---------------------------- | ---------------- | ---------- | ----------------------------------------------------------- | | `POST /v1/music/generations` | `music.generate` | ❌ No | Not reachable via the OpenAI SDK — call it over plain HTTP. | ### Audio | Operation | Infery SDK | OpenAI SDK | Notes / how to reach it otherwise | | -------------------------------- | ------------------------------ | ---------- | ------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------ | | `POST /v1/audio/speech` | `audio.speech.create` | ⚠️ Partial | Works for the common case — answers audio bytes. A fal/replicate-routed voice can instead answer a JSON 504 with a `job_id` in place of audio (still billed) — poll `GET /v1/images/jobs/{job_id}` for a signed download URL; the OpenAI SDK has no path for a JSON body where it expects bytes. | | `POST /v1/audio/transcriptions` | `audio.transcriptions.create` | ⚠️ Partial | Works, but timestamp\_granularities is dropped and srt/vtt/text come back as a JSON envelope. | | `POST /v1/audio/transformations` | `audio.transformations.create` | ❌ No | Not reachable via the OpenAI SDK — call it over plain HTTP. | ### 3D | Operation | Infery SDK | OpenAI SDK | Notes / how to reach it otherwise | | ------------------------- | ----------------- | ---------- | ----------------------------------------------------------- | | `POST /v1/3d/generations` | `threeD.generate` | ❌ No | Not reachable via the OpenAI SDK — call it over plain HTTP. | ### Files | Operation | Infery SDK | OpenAI SDK | Notes / how to reach it otherwise | | -------------------------------- | ---------------- | ---------- | --------------------------------- | | `POST /v1/files` | `files.create` | ✅ Yes | — | | `GET /v1/files` | `files.list` | ✅ Yes | — | | `GET /v1/files/{fileId}` | `files.retrieve` | ✅ Yes | — | | `GET /v1/files/{fileId}/content` | `files.content` | ✅ Yes | — | | `DELETE /v1/files/{fileId}` | `files.del` | ✅ Yes | — | ### Models & tools | Operation | Infery SDK | OpenAI SDK | Notes / how to reach it otherwise | | --------------------------------- | ----------------- | ---------- | ---------------------------------------------------------------------------------- | | `GET /v1/models` | `models.list` | ⚠️ Partial | Reachable, but the \_infery extension and client-side modality filtering are ours. | | `POST /v1/models/{slug}/estimate` | `models.estimate` | ❌ No | Not reachable via the OpenAI SDK — call it over plain HTTP. | | `GET /v1/tools` | `tools.list` | ❌ No | Not reachable via the OpenAI SDK — call it over plain HTTP. | ### Capabilities | Operation | Infery SDK | OpenAI SDK | Notes / how to reach it otherwise | | -------------------------------- | ------------------ | ---------- | ----------------------------------------------------------- | | `POST /v1/capabilities/{id}/run` | `capabilities.run` | ❌ No | Not reachable via the OpenAI SDK — call it over plain HTTP. | ### Workflows | Operation | Infery SDK | OpenAI SDK | Notes / how to reach it otherwise | | ------------------------------------- | ------------------------------ | ---------- | ------------------------------------------------------------------------------------------------------- | | `GET /v1/workflows` | `workflows.list` | ❌ No | Not reachable via the OpenAI SDK — call it over plain HTTP. | | `POST /v1/workflows` | `workflows.create` | ❌ No | Not reachable via the OpenAI SDK — call it over plain HTTP. | | `GET /v1/workflows/{id}` | `workflows.retrieve` | ❌ No | Not reachable via the OpenAI SDK — call it over plain HTTP. | | `PUT /v1/workflows/{id}` | `workflows.update` | ❌ No | Not reachable via the OpenAI SDK — call it over plain HTTP. | | `DELETE /v1/workflows/{id}` | `workflows.del` | ❌ No | Not reachable via the OpenAI SDK — call it over plain HTTP. | | `POST /v1/workflows/estimate` | `workflows.estimate` | ❌ No | Not reachable via the OpenAI SDK — call it over plain HTTP. | | `POST /v1/workflows/runs` | `workflows.runs.create` | ❌ No | mode: "stream" is refused here at compile time; workflows.runs.stream() is the SAME route, read as SSE. | | `GET /v1/workflows/runs/{id}` | `workflows.runs.retrieve` | ❌ No | Not reachable via the OpenAI SDK — call it over plain HTTP. | | `GET /v1/workflows/runs/{id}/logs` | `workflows.runs.logs` | ❌ No | Not reachable via the OpenAI SDK — call it over plain HTTP. | | `POST /v1/workflows/runs/{id}/cancel` | `workflows.runs.cancel` | ❌ No | Not reachable via the OpenAI SDK — call it over plain HTTP. | | `GET /v1/workflows/templates` | `workflows.templates.list` | ❌ No | Not reachable via the OpenAI SDK — call it over plain HTTP. | | `GET /v1/workflows/templates/{slug}` | `workflows.templates.retrieve` | ❌ No | Not reachable via the OpenAI SDK — call it over plain HTTP. | ## Reading "Partial" A "Partial" verdict means the OpenAI SDK's own method for that route sends a request this gateway accepts and can parse the normal response — but there is at least one real, non-hypothetical situation where the response diverges from what the OpenAI SDK expects (a different status code, a dropped parameter, an enveloped body where the SDK wants a bare one). The note says exactly what diverges and what to do about it. Treat "Partial" as "read the note before you ship," not as "basically yes."