Skip to content

v2.1.27

Latest

Choose a tag to compare

@cagostino cagostino released this 31 Oct 17:55
· 26 commits to main since this release
6c615d4

What's Changed

  • added cheapaimodel as a named LLM provider: OpenAI-compatible gateway at https://api.cheapaimodel.ai/v1, keyed by CHEAPAIMODEL_API_KEY with optional CHEAPAIMODEL_API_URL override, wired into get_litellm_response and resolve_model_provider, documented in README and the LLM responses guide (607eb38, 27ef14f, e5b753d)
  • added Firecrawl as a search_web provider via new search_firecrawl, which posts to the Firecrawl Search API v2 with FIRECRAWL_API_KEY and maps results to the standard title/link/content dicts (72e0d93, 4977281, e5b753d)
  • fixed litellm auto tool-call turns to add the final summary completion's token usage to reported usage totals, and to keep token accounting intact when the summary completion returns no usage data (2ceef44, d8e639f, e5b753d)
  • streaming litellm responses now snoop raw SSE lines for provider-reported usage/cost and emit a final usage event; serve.py MCP SSE and chat event streams request include_usage=True and report provider-reported cost via new provider_cost_from_usage helper (2ceef44)
  • fixed tool schema generation so array parameters (bare list, List, Optional[list]) always emit an items key, resolving OpenAI rejections of the generated function schemas (5f8a30b)
  • removed legacy console script entry points from setup.py (npc-init, npc-claude, npc-codex, npc-gemini, npc-opencode, npc-aider, npc-amp, npc-plugin since these now should be callable from npcsh (they're not quite yet but that will happen soon) ) (e5b753d)

PRs:

New Contributors

Full Changelog: v2.1.26...v2.1.27