v1.2.19 patches reasoning-content routing for thinking models, coerces MCP tool args to schema types, and allows multiple context providers of the same type in config.yaml.
Summary
Thinking model integration breaks without proper reasoning_content field mapping; tool schema mismatches halt execution silently. These fixes remove friction when chaining local inference (Ollama) with reasoning models (Kimi, Gemini 3) across the same config.
Why it matters
Thinking model integration breaks without proper reasoning_content field mapping; tool schema mismatches halt execution silently. These fixes remove friction when chaining local inference (Ollama) with reasoning models (Kimi, Gemini 3) across the same config.
Implementation verdict
Drop-in patch for existing Continue users; no migration required. If you're routing reasoning-delta through Kimi or Gemini 3, upgrade immediately—the previous version silenced reasoning output. For MCP tool integration, this fixes schema validation errors on OpenAI Adapter calls. Worth upgrading now if you use thinking models or multiple Ollama contexts in a single config.
Sources
Dev Signal
Get briefs like this in your inbox — free, 3x a week.
100+ sources compressed into one 4-minute read. Ranked, cited, implementation-ready.