SDK
5/20/26Less than 1 minute
PulseNeko Bridge speaks the three major protocols. Native SDKs work unmodified — swap base_url and api_key.
Protocols
| Protocol | Base URL | SDKs |
|---|---|---|
| OpenAI Chat / Responses | https://bridge.pulseneko.com/v1 | openai, @ai-sdk/openai, etc. |
| Anthropic Messages | https://bridge.pulseneko.com | anthropic |
| Gemini native | https://bridge.pulseneko.com/gemini | google-genai, @google/genai |
Cross-language: official SDKs in Python / Node / Go / Java / Rust. Anything else: plain HTTP + Bearer auth.
Frameworks
| Framework | How |
|---|---|
| LangChain | Set LLM's base_url / openai_api_base to Bridge |
| LlamaIndex | Same idea (api_base) |
| Vercel AI SDK | Use @ai-sdk/openai with custom baseURL |
By language
- OpenAI-compatible — Python / Node / cURL
- Anthropic — Claude family
- Gemini — Google native
- LangChain — framework integration
General notes
- Timeouts — default SDK timeout is usually 60s; long-context / reasoning models warrant 300s+
- Streaming — all SDKs support
stream=True; Bridge passes SSE through - Retries — let the SDK handle transient
502/504(max_retries=2) - Concurrency — client instances are thread-safe and shareable