feat: bootstrap jetson-ai-ollama from agent-swarm template
Some checks failed
gates / quality (push) Has been cancelled

Leader-Bootstrap (Planung) aus Template + Jetson-Meta-Prompt:
- ROADMAP-Board W1-W6 als Abhaengigkeits-DAG (Ready-Set = {W1})
- STATE mit disjunkter Datei-Ownership W2-W6
- CONVENTIONS Projekt-Vertrag (Profile, Pfade, Ports, Daemon-Env, num_ctx, DVC, Scope)
- manifests/requirements.md = kalt verifizierbares Review-Manifest (R-01..R-10)
- REVIEW.md Kalt-Review-Einstieg; ADR 0002 (Ollama-nativ, kein Swapper/Gateway)
- session-prompts/W1..W6 dispatch-fertig; .gitignore um Jetson-Ignores erweitert
- examples/walkthrough entfernt (Projekt hat eigenes gefuelltes Board)

Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
This commit is contained in:
2026-06-03 06:55:39 +02:00
commit efee259d01
43 changed files with 1563 additions and 0 deletions

20
manifests/requirements.md Normal file
View File

@@ -0,0 +1,20 @@
# Requirements-Manifest (kalt verifizierbar)
Eine Zeile je Erfolgskriterium — der **kalt lesbare Review** (siehe `REVIEW.md`).
Der Reviewer prüft Spalte für Spalte; `done`-Zeilen müssen ihren Prüfbefehl bestehen.
`scripts/gate-doc-drift.sh` erzwingt die Existenz der Pfade in `done`/`in_progress`-Zeilen.
| ID | Anforderung | Datei(en) | Test/Prüfbefehl | Status |
|----|-------------|-----------|------------------|--------|
| R-01 | Vertrag zuerst: CONVENTIONS + disjunkte Ownership | `CONVENTIONS.md`, `STATE.md` | `grep -q qwen-light CONVENTIONS.md` | done |
| R-02 | Drei Profile über Modelfiles (light/gemma/heavy) | `ollama/Modelfile.light`, `ollama/Modelfile.gemma`, `ollama/Modelfile.heavy` | `pytest tests/test_modelfiles.py` | todo |
| R-03 | q8_0-KV + Flash Attention + MAX_LOADED_MODELS=1 | `ollama/systemd/ollama.service.d/override.conf` | `grep -Eq 'q8_0' ollama/systemd/ollama.service.d/override.conf` | todo |
| R-04 | num_ctx je Profil explizit (light/gemma 16384, heavy 8192) | `ollama/Modelfile.light`, `ollama/Modelfile.heavy` | `grep -q 'num_ctx' ollama/Modelfile.light` | todo |
| R-05 | GPU-Nutzung verifiziert; Ollama nur LAN + Firewall | `scripts/setup-jetson.sh` | `bash -n scripts/setup-jetson.sh` | todo |
| R-06 | Metrik-Endpoint über Exporter (`:8000/metrics`) | `exporter/systemd/ollama-exporter.service` | `pytest tests/test_endpoints.py` | todo |
| R-07 | Keine Secrets; `.env.example` nur Platzhalter | `.env.example` | `grep -q 'JETSON_LAN_IP=' .env.example` | todo |
| R-08 | DVC = Modell-Source-of-Truth; build reproduzierbar | `scripts/build-models.sh`, `.dvc/config` | `bash -n scripts/build-models.sh` | todo |
| R-09 | Client + CLI funktionsfähig (OpenAI-kompatibel) | `client/python/jetson_ai/client.py`, `client/cli/ai.py` | `pytest tests/` | todo |
| R-10 | Kalt verifizierbar; Tests kodieren Kriterien | `REVIEW.md`, `tests/test_endpoints.py` | `pytest tests/ -q` | todo |
Status: `todo` · `in_progress` · `done` · `drift` (vom Audit markiert).