orchestrata_ultimate.llm
Reference › orchestrata_ultimate › llm
Orchestrata LLM client — async facade for volk brains.
The llm package is the single way volks reach a language model. A volk calls call_llm, call_reasoner or call_with_tools, and the package decides which provider and model answer, sends the request through the shared provider manager, retries transient failures and reports final failure with a small exception vocabulary. The hand-written story is llm; the base agent that calls it is base.
The parts: config maps volk aliases to providers and models from the YAML profile selected by SPIKE_PROFILE, with no hardcoded defaults, and can escalate a model to the reasoning tier; client holds the call functions, adapter registration, retry and conversation capture; _cassette provides record and replay of exchanges for tests; exceptions defines LLMClientError and LLMProviderUnavailableError; init re-exports the public names.
Control flows from a volk's call into client, which asks config for the model, builds a ProviderRequest and passes it to the ProviderManager of the providers package. Start reading at client.call_llm, then config.VolkConfig.
Curated narrative
The hand-written story of this package: orchestrata_ultimate.llm.
5 modules, 1,894 lines of code.
Modules
| Module | Summary | LOC |
|---|---|---|
__init__ |
Orchestrata LLM client — async facade for volk brains. | 40 |
_cassette |
LLM record/replay cassette store. | 325 |
client |
Async LLM client for orchestrata volk brains. | 1077 |
config |
Per-volk LLM provider and model configuration. | 429 |
exceptions |
LLM client exceptions. | 23 |
Dependency diagram
