Function llm_request_impl 

pub fn llm_request_impl(
    lua: &Lua,
    args: (String, Option<Table>),
) -> Result<Table>

Executes an LLM chat request. Called from capability-gated context.

Arguments (from Lua)

• prompt - User message text
• opts - Optional table:
  • provider - “ollama” (default), “openai”, “anthropic”
  • base_url - Provider base URL (default per provider)
  • model - Model name (default per provider)
  • api_key - API key (falls back to env var)
  • system_prompt - System prompt text
  • session_id - Session ID for multi-turn (nil = new session)
  • temperature - Sampling temperature
  • max_tokens - Max completion tokens
  • timeout - Per-request timeout in seconds (default: 120)
  • overall_timeout - Wall-clock timeout in seconds for the entire resolve loop (default: nil = no limit). When set, the loop aborts with error_kind = "overall_timeout" if the deadline is exceeded.

Returns (Lua table)

• ok - boolean
• content - Response text (when ok=true)
• model - Model name from response
• session_id - Session ID (new or existing)
• error - Error message (when ok=false)
• error_kind - Error classification