ACE
A context engine and retrieval infrastructure for task-conditioned retrieval, source ranking and pruning, context budgets, and citation-preserving assembly across AI workspaces and agent workflows.
Treats context as infrastructure: ranks, prunes, and routes sources under explicit budgets so agents see only what is strictly necessary for the current step — with citations preserved.
Problem Space
Most retrieval systems fail by providing too much irrelevant information: token waste, model confusion, missing provenance, stale sources, and no respect for who is allowed to see what.
System Direction
ACE combines task-conditioned retrieval with source ranking and pruning under explicit context budgets. Retrieval is hierarchical and both structural and semantic; citations are preserved through assembly; recency is handled explicitly; and authority-aware ACLs plus workspace and execution context bound what any task may retrieve.
Public Capabilities
- 01Task-conditioned retrieval
- 02Source ranking and pruning
- 03Context budgets
- 04Hierarchical retrieval
- 05Structural plus semantic retrieval
- 06Citation preservation
- 07Recency handling
- 08Authority-aware ACLs
- 09Workspace and execution context
ACE is preparing for open source release. The repository will be made public when the initial release is ready.
What Is Not Disclosed
Private implementation details, security-sensitive internals, and unreleased runtime architecture are intentionally not disclosed.