nousresearch / hermes-agent
Code health
This codebase scores 6.1 out of 10 for code health, which we rate fair. It also scores maintainability 7.4 and static performance 9.5 out of 10. The three are scored separately and never blended into one number. Risk is concentrated, as it usually is: 1,424 of 8,758 files are git hotspots, and they average 3.7 — which is where the fixes pay off most.
Do next
2 things worth doing this quarter.
Worth planning
Update the decision “Fan out passthrough button presses to all live gateway sessions”: the code it governs has moved on
Agents and reviewers still follow it, so they are steered by a rule the code no longer keeps.
- status
- Accepted, drifting
Small effort, worth a checkDelete 86 unused symbols and files (1,380 lines)
Nothing in the graph reaches them, across 58 files; every reader and agent pays to skip them.
- symbols
- 86
- lines
- 1,380
- safe to delete
- Yes(inferred)
Medium effort
Improve what Repowise can see
Add a test coverage report
Without one, Repowise cannot tell tested code from untested code, so every test-related action says “unknown”.
- line coverage
- Unknown, no report
Small effort
- Lines of code
- 2.1M
Recent activity
All commitsCommits
- 13ce0c5cfmt(js): `npm run fix` on merge (#89914)1mo ago
- 1a19fedbfix(desktop): Bot Mode chats no longer paint blank when switching between running bots1mo ago
- ac3d7ddafix(desktop): skill hub installs on non-default profiles no longer 404, and failed installs surface1mo ago
- dc77f2c8Revert "ci: force workflow re-parse after zero-job dispatch window"1mo ago
- 43e67f1fci: move orchestrator to ci.yaml — ci.yml workflow identity is wedged (0-job startup_failure on every dispatch; identical content dispatches fine under a new path, proven by probe PR #89894)1mo ago
- ab173e26ci: force workflow re-parse after zero-job dispatch window1mo ago
Decisions
- Fan out passthrough button presses to all live gateway sessionsactive · 1mo ago
- Track auxiliary model usage per task in session accounting without double-counting sessions countersproposed · 1mo ago
- Fix Anthropic OAuth file per-profile and close missing port-binding platform set gapsproposed · 1mo ago
- Guard named custom provider routing and fix named-custom key assignmentproposed · 1mo ago
- Ensure Anthropic messages start with a user role after compactionproposed · 1mo ago
- Coerce empty/whitespace-only Anthropic text blocks to avoid HTTP 400 and session wedgingproposed · 1mo ago
Everything we found
48,817 open
Every detector's open count by area, each with its worst example. An inventory, not a to-do list: the actions above are the part worth doing.
- worst item is critical severity. Code health36,587Untested hotspot · gateway/run.py1260 findings · 10.2 deducted
- worst item is medium severity. Refactoring5,930usage_model_from_account+1.5 health, S effort · agent/billing_usage.py
- worst item is high severity. Decisions1,465Fan out passthrough button presses to all live gateway sessions1 drifting from the code · 44 awaiting review · 1420 hotspots ungoverned
- worst item is medium severity. Ownership4,733agent/relay_llm.py93% single-owner concentration, 40 commits in 90d
- worst item is low severity. Dead code102get_accounting_contextUnreachable function (2 lines)
Composition
Open the graphExplore this codebase
- Docs9,727 pages across 24 modules
- ChatAsk this codebase a question and get an answer with its sources.
- Files8,758 files with per-file docs, health and history
- ArchitectureDependency graph, layers, and 236 entry points
- Code healthPer-file scores, 43,909 open findings, coverage and refactoring targets
- Knowledge graphEntities, communities, and the paths between them
- Change couplingFiles that keep changing together, mined from commit history
- CommitsChange-risk ranked history with agent provenance
- ContributorsBus factor, per-file maintainers, and the human/agent split
- StatsSize class, origin, lifetime churn, rhythm and records
- CostsWhat indexing this snapshot cost, by model and by run.