ReddRadar
Agent skill
Convex Agents Context
Customizes what information the LLM receives for each generation. Use this to control message history, implement RAG context injection, search across threads, and provide custom context.
Install this agent skill to your Project
npx add-skill https://github.com/majiayu000/claude-skill-registry/tree/main/skills/data/convex-agents-context
SKILL.md
Purpose
By default, the Agent includes recent messages as context. This skill covers customizing that behavior for advanced patterns like cross-thread search, memory injection, summarization, and filtering.
When to Use This Skill
- Limiting context window to prevent token overflow
- Searching across multiple threads for relevant context
- Injecting memories or user profiles into every prompt
- Summarizing long conversations before continuing
- Filtering out sensitive or irrelevant messages
- Adding few-shot examples to guide LLM
Configure Default Context Options
const myAgent = new Agent(components.agent, {
name: "My Agent",
languageModel: openai.chat("gpt-4o-mini"),
contextOptions: {
recentMessages: 50,
excludeToolMessages: true,
searchOptions: {
limit: 10,
textSearch: true,
vectorSearch: false,
},
},
});
Override Context Per Call
export const generateWithCustomContext = action({
args: { threadId: v.string(), prompt: v.string() },
handler: async (ctx, { threadId, prompt }) => {
const result = await myAgent.generateText(
ctx,
{ threadId },
{ prompt },
{
contextOptions: {
recentMessages: 20,
searchOptions: {
limit: 5,
textSearch: true,
vectorSearch: true,
},
},
}
);
return result.text;
},
});
Search Across Threads
export const generateWithCrossThreadContext = action({
args: { threadId: v.string(), userId: v.string(), prompt: v.string() },
handler: async (ctx, { threadId, userId, prompt }) => {
const result = await myAgent.generateText(
ctx,
{ threadId, userId },
{ prompt },
{
contextOptions: {
searchOtherThreads: true,
searchOptions: {
limit: 15,
textSearch: true,
vectorSearch: true,
},
},
}
);
return result.text;
},
});
Custom Context Handler
Completely customize context:
const myAgent = new Agent(components.agent, {
name: "My Agent",
languageModel: openai.chat("gpt-4o-mini"),
contextHandler: async (ctx, args) => {
const userMemories = await getUserMemories(ctx, args.userId);
const examples = getExamples();
return [
...userMemories,
...examples,
...args.search,
...args.recent,
...args.inputMessages,
];
},
});
Fetch Context Manually
Get context without calling LLM:
import { fetchContextWithPrompt } from "@convex-dev/agent";
export const getContextForPrompt = action({
args: { threadId: v.string(), prompt: v.string() },
handler: async (ctx, { threadId, prompt }) => {
const { messages } = await fetchContextWithPrompt(ctx, components.agent, {
threadId,
prompt,
contextOptions: {
recentMessages: 20,
searchOptions: { limit: 10, textSearch: true },
},
});
return messages;
},
});
Key Principles
- Default context is sufficient: Most use cases work with defaults
- Search improves relevance: Enable for long conversations
- userId required for cross-thread: Provide when searching multiple threads
- Context handlers are powerful: Use for memories, examples, special formatting
- Recent messages take precedence: Used after search in context order
Next Steps
- See rag for knowledge base context injection
- See fundamentals for agent setup
- See rate-limiting for token management
Recommended Agent Skills
Expand your agent's capabilities with these related and highly-rated skills.
majiayu000/claude-skill-registry
agent-ops-spec
Manage specification documents in .agent/specs/. Use when user provides requirements, acceptance criteria, or feature descriptions that need to be tracked and validated against implementation.
majiayu000/claude-skill-registry
agent-ops-state
Maintain .agent state files. Use at session start, after meaningful steps, and before concluding: read/update constitution/memory/focus/issues/baseline consistently.
majiayu000/claude-skill-registry
agent-ops-spec
Manage specification documents in .agent/specs/. Use when user provides requirements, acceptance criteria, or feature descriptions that need to be tracked and validated against implementation.
majiayu000/claude-skill-registry
agent-ops-testing
Test strategy, execution, and coverage analysis. Use when designing tests, running test suites, or analyzing test results beyond baseline checks.
majiayu000/claude-skill-registry
agent-ops-testing
Test strategy, execution, and coverage analysis. Use when designing tests, running test suites, or analyzing test results beyond baseline checks.
majiayu000/claude-skill-registry
agent-ops-state
Maintain .agent state files. Use at session start, after meaningful steps, and before concluding: read/update constitution/memory/focus/issues/baseline consistently.
Didn't find tool you were looking for?