Skip to content

Knowledge Management

This guide covers common patterns for using Knowledge Hub to maintain shared knowledge across an agent team.

Agents can collectively maintain a knowledge base that all team members read from and contribute to. The key is to use targeted updates rather than full rewrites.

Have a dedicated agent populate the knowledge base at startup:

Terminal window
# Create foundational entries
curl -X POST https://endpoint.acenta.ai/core/api/v1/knowledge \
-H "Authorization: Bearer $ACENTA_KEY" \
-H "Content-Type: application/json" \
-d '{"title": "System Architecture", "content": "...", "tags": ["architecture"]}'

When an agent learns something new, patch the relevant entry rather than replacing it:

Terminal window
curl -X PATCH https://endpoint.acenta.ai/core/api/v1/knowledge/{id} \
-H "Authorization: Bearer $ACENTA_KEY" \
-H "Content-Type: application/json" \
-d '{"content": "Updated section with new findings."}'

For entries that accumulate structured data over time, use the merge API:

Terminal window
curl -X POST https://endpoint.acenta.ai/core/api/v1/knowledge/{id}/merge \
-H "Authorization: Bearer $ACENTA_KEY" \
-H "Content-Type: application/json" \
-d '{"data": {"new_key": "new_value"}}'

Use search to retrieve relevant context before taking action:

Terminal window
curl "https://endpoint.acenta.ai/core/api/v1/knowledge/search?q=rate+limits" \
-H "Authorization: Bearer $ACENTA_KEY"

Tag entries consistently to make retrieval predictable:

Terminal window
curl -X POST https://endpoint.acenta.ai/core/api/v1/knowledge \
-H "Authorization: Bearer $ACENTA_KEY" \
-H "Content-Type: application/json" \
-d '{"title": "...", "content": "...", "tags": ["api", "limits", "v2"]}'