LUMOS

LUMOS

LUMOS

Give AI the right context before it touches your code.

Lumos finds the files, proof chains, and tests your coding agent needs before it writes a single line.

Open LumosSee how it works

Graph-native context

One request becomes a complete trail to the code that matters.

Lumos follows real repository relationships, then returns only the files, tests, and evidence a coding agent needs.

Live repository graph

See the evidence assemble around one request.

Lumos starts with the issue, follows HydraDB relationships, and lights up the smallest defensible path to the edit.

Relationship trail
Definition + coverage
Focused context
12 signals
Safety coverage
20 checks
Selected evidence chainSurrounding context
Evidence mapConnected

The evidence route

From request to ready-to-edit context.

Four graph-native actions turn an issue into a handoff an agent can inspect and use.

  1. 01

    Resolve

    Turn the issue into concrete files and symbols.

  2. 02

    Trace

    Walk the HydraDB graph through code relationships.

  3. 03

    Prove

    Attach inspectable evidence and relevant tests.

  4. 04

    Handoff

    Send a compact context package to the coding agent.

Counterfactual proof

See exactly what the graph changed.

Word search ranked the change area third. HydraDB connected the intended behavior to its safety signal and promoted the right context to first.

Word search
#3
Lumos
#1
Graph walk
6 ms
Why HydraDB changed the answer
Lexical baseline#3

Names match several nearby files.

Evidence-backed rank#1

The graph connects the symbol to its covering test.

  1. Named behaviorTemplate joining
  2. Change surfaceOutput escaping boundary
  3. Safety signalRendered output contract

Without HydraDB, there is no reverse call chain and no test-impact path to justify the promotion.

The best context is not the largest. It is the smallest set of evidence an agent can trust.

Lumos turns repository structure into a clear starting point for every change.

Explore the documentation