Add multi-pattern orchestrator with plan-then-execute and supervisor (#739)

Introduce an agent orchestrator service that supports three
execution patterns (ReAct, plan-then-execute, supervisor) with
LLM-based meta-routing to select the appropriate pattern and task
type per request. Update the agent schema to support
orchestration fields (correlation, sub-agents, plan steps) and
remove legacy response fields (answer, thought, observation).
This commit is contained in:
cybermaggedon 2026-03-31 00:32:49 +01:00 committed by GitHub
parent 7af1d60db8
commit 849987f0e6
No known key found for this signature in database
GPG key ID: B5690EEEBB952194
21 changed files with 3006 additions and 172 deletions

View file

@ -56,6 +56,7 @@ Homepage = "https://github.com/trustgraph-ai/trustgraph"
[project.scripts]
agent-manager-react = "trustgraph.agent.react:run"
agent-orchestrator = "trustgraph.agent.orchestrator:run"
api-gateway = "trustgraph.gateway:run"
chunker-recursive = "trustgraph.chunking.recursive:run"
chunker-token = "trustgraph.chunking.token:run"