Set Up GigaChat API Access

GigaChat Setup handles everything needed before any GigaChat application code should be written: creating a project on developers.sber.ru, turning a client_id/client_secret pair into the base64 authorization key GigaChat expects, choosing the correct scope for the account type, and configuring TLS certificates for the Russian Trusted Root CA that GigaChat's endpoints require. It also documents the base environment variables shared across the gigachat SDK, langchain-gigachat, and gpt2giga, so credentials are set once and reused consistently.

Use it whenever a task starts with something like "how do I connect to GigaChat?" - registration, first-time credential setup, or fixing auth, TLS, or scope errors on an existing setup. It explicitly stays out of implementation-level concerns: it will not explain chat payload design (that's gigachat-sdk-chat), function calling (gigachat-sdk-functions), LangChain wrappers (langchain-gigachat), or proxy routing (gpt2giga) - each of those has its own skill, and this one only hands off to them once connectivity is confirmed working.

The workflow is sequential: guide registration if credentials don't exist yet; generate the base64 authorization key from client_id:client_secret if they do; confirm the account type to pick between GIGACHAT_API_PERS, GIGACHAT_API_B2B, or GIGACHAT_API_CORP scopes; get TLS verification working correctly before ever suggesting the escape hatch of verify_ssl_certs=False; and verify the whole chain with a minimal call like get_models() before touching application code. Failures are triaged in a fixed order: TLS, then auth, then scope, then rate limit, then content policy. Two reference files carry the deeper detail: references/auth-and-certs.md for setup, environment variables, token flow, and certificates, and references/troubleshooting.md for diagnosing common failures.

The skill's environment-variable and certificate guidance is written to be shared across every GigaChat access path in the ecosystem - the official gigachat Python SDK, the langchain-gigachat integration, and the gpt2giga OpenAI-compatible proxy - so credentials configured once via this skill work the same way regardless of which of those a project later uses. Its default posture prefers fixing the CA bundle over disabling TLS verification, and prefers environment variables over hardcoded secrets for SDK or proxy usage.

Developers setting up GigaChat access for the first time, or debugging an existing integration that's failing with auth, TLS/SSL, or scope errors, before moving on to the chat, function-calling, LangChain, or proxy skill that actually implements their use case.