Module api 

LLM API entry points and re-exports. The transport layer, request/ response parsing, auth, context-window discovery, and option/result types each live in their own submodule under self; this file only wires them together and hosts the vm_call_llm_full* chat entry points that provider-specific completion / agent paths dispatch into.

Modules

telemetry_source

Wire-format identifiers for ProviderTelemetry::source. Keep these in sync with the matching strings in docs/src/observability/* and the The host eval aggregator.

Structs

OllamaPsModel

Loaded-model snapshot from Ollama’s /api/ps. Shared with the CLI’s harn local family so we don’t duplicate the response shape.

OllamaReadinessOptions

OllamaReadinessResult

OllamaRuntimeSettings

OllamaWarmupResult

ProviderTelemetry

Provider-side timing and runtime accounting captured per LLM call.

Constants

HARN_OLLAMA_KEEP_ALIVE_ENV

HARN_OLLAMA_NUM_CTX_ENV

OLLAMA_DEFAULT_KEEP_ALIVE

OLLAMA_DEFAULT_NUM_CTX

OLLAMA_HOST_ENV

Functions

fetch_provider_max_context

Fetch the server-reported maximum context length for a given model, if available. Caches results per (base_url, model_id) so we only pay the discovery cost once per session.

normalize_ollama_keep_alive

Public wrapper around the internal keep-alive parser, used by callers (CLI flags, host bridges) that want the same normalization Harn applies to environment overrides.

ollama_readiness

ollama_runtime_settings_from_env

warm_ollama_model

warm_ollama_model_with_settings