Navigate GigaChat Ecosystem

GigaChat Navigation is a routing skill for the GigaChat ecosystem: it decides whether a task should be built on the official gigachat Python SDK, on langchain-gigachat, or on gpt2giga (the OpenAI-compatible proxy), based on the integration style the task actually needs rather than which API buzzwords appear in the request. It also tracks which capabilities and limitations across these three layers are actually verified, and points to the right next skill or reference file once a layer is chosen.

Use it first, before any implementation work, whenever it's unclear which GigaChat integration layer fits the task at hand. It is explicitly a decision and hand-off point, not an implementation skill itself - it does not replace the dedicated skills for SDK chat, SDK files/functions, LangChain usage, or proxy work, and should always route into one of them rather than trying to do the implementation directly.

The routing logic is a short decision tree: if the user is writing a native Python integration, prefer gigachat; if the application already uses LangChain primitives, use langchain-gigachat; if the client needs to keep an OpenAI-compatible or Anthropic-compatible SDK unchanged, use gpt2giga. When more than one layer could technically work, the guidance is to pick the simplest path that's been locally verified rather than the most feature-complete one in theory. One cross-cutting constraint carried over into every layer: file-based requests should stay limited to a single modality per request. Two reference files back the routing: references/decision-rules.md for the routing logic itself, and references/feature-matrix.md for which features are verified as working on each layer.

This skill is the entry point to a small family of GigaChat-focused skills - the gigachat-sdk-* skills for direct SDK usage (chat, files, embeddings, functions), langchain-gigachat for LangChain-based chains and RAG, and gpt2giga for OpenAI-compatible proxying - and its main job is naming which of those to use next and stating explicitly why the others aren't the default for this particular task.

Developers or agents starting a new GigaChat integration who aren't sure yet whether to reach for the raw SDK, a LangChain wrapper, or an OpenAI-compatible proxy, and want a quick, verified answer before committing to an implementation path.