tesserakit-app

The Tessera app: run the whole hub over a project and get one browsable dashboard.

tessera-app is the unifying surface over the job packs. It detects which packs apply to a project, runs them, and renders all their artifacts into a single self-contained HTML dashboard. No server, no build step, no extra dependencies.

It is a CLI-only plugin: it registers commands under tessera.commands and orchestrates JobPacks, but is not itself a JobPack. (This is the first real use of the deliberate split between the tessera.commands and tessera.jobpacks entry-point groups.)

Commands

tessera detect --input .              # show which packs apply, without running
tessera run --input . --output run    # run applicable packs + build dashboard
tessera dashboard --input run         # (re)build the HTML dashboard from a run

What run does

1. Detects applicable packs by inspecting the project:
   • .prompt.md / PROMPT.md → prompts
   • SKILL.md → skills
   • .recipe.md / RECIPE.md → recipes
   • .curl / curl .sh → api
   • corpus/ + a queries.* file → rag
   • any .csv → evals (task generic)
   • source files or a dependency manifest → repo
2. Runs each applicable pack into output/<pack>/, continuing past any pack that fails.
3. Writes run_manifest.json summarizing record/finding counts and artifacts.
4. Builds output/index.html: a dashboard with headline cards and every pack's reports, rendered from Markdown to HTML in the browser-ready file.

Notes

• The orchestrator never raises on a single pack's failure; it records the error and moves on, so one bad input does not sink the whole run.
• The dashboard is fully self-contained (inline CSS, no JS, no external assets) so it can be committed, emailed, or served as a static file.
• For full control over a single pack (e.g. a specific eval task type or column overrides), use that pack's own command (tessera evals compile ...).