Convex Agents Context

You can control LLM context by customizing message history, enabling cross-thread search, and injecting memories to tailor the exact information the LLM receives for each generation call.

Yes, cross-thread context search allows you to retrieve relevant history across multiple conversations, ensuring agents have the necessary background information from different threads.

You can inject memories, profiles, or examples into the LLM context to guide the agent's behavior consistently over time, tailoring prompts for specific tasks and maintaining continuity.

Yes, you can configure per-call context options to override default settings, allowing you to tailor prompts and context dynamically for specific tasks within multi-thread conversations.

The skill supports manual context fetch, allowing you to retrieve and customize agent context explicitly for memory-aware prompts in advanced multi-thread conversation setups.

Yes, context summarization is supported alongside memory injection and cross-thread search, enabling you to manage and condense the information fed to the LLM for smarter agent responses.