Output Files
All output is written relative to the working directory — the directory you run otx from, or the directory the web app was started in.
Decomposition Output
Output/Decomposition/<project>/
| File | Phase | Description |
|---|---|---|
00-index.md | 0 | Architectural overview and key design decisions |
01-structural-survey.md | 1 | Directory tree, entry points, config files, dependencies, toolchain |
02-initialization-flow.md | 2 | Startup sequence, hooks, shutdown, async tasks |
03-01-<name>.md … 03-N-<name>.md | 3 | Component specifications (one file per cluster) |
04-data-formats.md | 4 | File formats, protocols, public API surfaces |
05-reimplementation-checklist.md | 5 | Ordered checklist with acceptance criteria and security analysis |
06-composition-inventory.md | 6 | Flat inventory catalogue — source for database import |
07-ethos.md | 7 | Style fingerprint and coding standards guide |
inventory.json | post-6 | Per-project inventory export in JSON |
job.json | ongoing | Job state snapshot — used for resume-on-failure |
The number of 03-NN-*.md files varies by project. Phases 4 and 5 always use 04- and 05- prefixes regardless of how many component files Phase 3 produced.
Composition Output
Output/Composition/<name>/
| File | Description |
|---|---|
compose-output.md | Full compose or transmute output — architectural design document |
Implementation Output
Output/Implementation/<project>/
All files and directories written by the Implement pipeline. The structure is determined by the AI agent based on the spec.
Database
DB/opentransmute.db
SQLite database containing:
- Decomposed projects (name, source URL, decomposed timestamp)
- Inventory items (all seven categories, structured fields, raw markdown)
- Phase outputs (content of each completed phase file, for web app preview)
- Compose job records
- User settings singleton
The database is created automatically on first run. It is shared between the web app and the CLI when both are run from the same working directory.
CLI Settings
~/.opentransmute/settings.json
User-level settings file for the CLI. Contains orchestrator preference, model names, endpoint URL, turn/token limits, and user ethos text. API keys are never written here.
Exporting the Inventory
Via web app — Inventory screen → Export current view → downloads filtered results as JSON.
Via API:
GET /api/inventory/export?projectId=<guid> # one project
GET /api/inventory/export # all projects
Export format:
{
"project": {
"name": "my-project",
"source": "https://github.com/org/repo",
"decomposedAt": "2026-04-17T10:00:00Z"
},
"items": [
{
"id": "...",
"category": "Algorithm",
"name": "Token Bucket Rate Limiter",
"summary": "...",
"details": {
"inputs": "...",
"outputs": "...",
"pseudocode": "..."
},
"rawMarkdown": "..."
}
]
}