chore(llm): refresh defaults to match deployed Jetson backend (#2)

.gitea/workflows/cd.yml

@@ -0,0 +1,46 @@
+# CD — rebuild and restart ci-ai-bot on the Pi whenever bot/ or scripts/ change.
+#
+# How it works (Docker-out-of-Docker):
+#   1. Job runs in docker:cli on the Pi runner (eulernest label).
+#   2. Mounts the HOST docker socket → builds the new image directly in the host
+#      Docker daemon (no registry push needed — image is already local).
+#   3. Mounts /mnt/external:ro → runs `docker compose up --force-recreate`
+#      which picks up the freshly built image and recreates the container.
+#
+# Prerequisite (one-time):
+#   - Runner config must list /var/run/docker.sock and /mnt/external in valid_volumes
+#     (set in ansible-eulernest roles/gitea_runner/templates/config.yaml.j2).
+#   - Set org variable JETSON_OLLAMA_URL in conformallab → Actions → Variables:
+#     value = http://192.168.178.103:11434/v1
+name: CD — ci-ai-bot
+
+on:
+  push:
+    branches: [main]
+    paths:
+      - "bot/**"
+      - "scripts/**"
+
+jobs:
+  deploy:
+    runs-on: eulernest
+    container:
+      image: docker:27-cli
+      volumes:
+        - /var/run/docker.sock:/var/run/docker.sock
+        - /mnt/external:/mnt/external:ro
+
+    steps:
+      - uses: actions/checkout@v4
+
+      - name: Build ci-ai-bot image
+        # Build from repo root so bot/Dockerfile can COPY scripts/
+        run: docker build -t ci-ai-bot:latest -f bot/Dockerfile .
+
+      - name: Recreate ci-ai-bot container
+        # --force-recreate: always replace the container even if config is unchanged,
+        # so the freshly built image is picked up.
+        run: |
+          docker compose \
+            -f /mnt/external/docker-compose.yml \
+            up -d --no-deps --force-recreate ci-ai-bot

.gitea/workflows/reusable-ai-review.yml

@@ -30,7 +30,8 @@ jobs:
         env:
           AI_PROVIDER: ollama
           AI_MODEL: coder
-          AI_BASE_URL: http://192.168.178.103:11434/v1
+          # JETSON_OLLAMA_URL: org variable in conformallab → Actions → Variables
+          AI_BASE_URL: ${{ vars.JETSON_OLLAMA_URL || 'http://192.168.178.103:11434/v1' }}
           GITEA_TOKEN: ${{ github.token }}
           GITEA_API: https://git.eulernest.eu/api/v1
         run: python /opt/ci-ai/scripts/ai_review.py

.gitea/workflows/reusable-commands.yml

@@ -22,7 +22,8 @@ jobs:
     env:
       AI_PROVIDER: ollama
       AI_MODEL: coder
-      AI_BASE_URL: http://192.168.178.103:11434/v1
+      # JETSON_OLLAMA_URL: org variable in conformallab → Actions → Variables
+      AI_BASE_URL: ${{ vars.JETSON_OLLAMA_URL || 'http://192.168.178.103:11434/v1' }}
       GITEA_TOKEN: ${{ github.token }}
       GITEA_API: https://git.eulernest.eu/api/v1
     steps:

.gitea/workflows/reusable-security.yml

@@ -73,7 +73,8 @@ jobs:
         env:
           AI_PROVIDER: ollama
           AI_MODEL: coder
-          AI_BASE_URL: http://192.168.178.103:11434/v1
+          # JETSON_OLLAMA_URL: org variable in conformallab → Actions → Variables
+          AI_BASE_URL: ${{ vars.JETSON_OLLAMA_URL || 'http://192.168.178.103:11434/v1' }}
           GITEA_TOKEN: ${{ github.token }}
           GITEA_API: https://git.eulernest.eu/api/v1
         run: |

README.md

@@ -125,15 +125,16 @@ Provider-Wechsel = **nur Env**, kein Code-Change (`scripts/llm.py`). Anthropic-V
 
 ## Konfiguration (Env-Vars)
 
-> Hinweis: Die **Code-Defaults in `llm.py` sind noch der alte Stand** (Anthropic, Jetson-IP
+> Hinweis: Die **Code-Defaults in `llm.py` entsprechen jetzt dem Jetson-Betrieb**
-> `.20`, `qwen-light`). Im Betrieb überschreiben Workflows + Bot diese per Env mit den
+> (`ollama` / `coder` / `.103`). Workflows + Bot setzen die Werte trotzdem explizit
-> Jetson-Werten (`ollama` / `coder` / `.103`). Ein Refresh der Defaults ist vorgemerkt.
+> per Env — ein Provider-Wechsel (z.B. zurück auf `anthropic`) bleibt damit ein
+> reiner Env-Change, keine Code-Änderung.
 
 | Var | Default (`llm.py`) | Im Betrieb |
 |---|---|---|
-| `AI_PROVIDER` | `anthropic` | **`ollama`** |
+| `AI_PROVIDER` | `ollama` | `ollama` |
-| `AI_MODEL` | `claude-3-5-haiku-latest` / `qwen-light` | **`coder`** |
+| `AI_MODEL` | `coder` (ollama) / `claude-3-5-haiku-latest` (anthropic) | `coder` |
-| `AI_BASE_URL` | `http://192.168.178.20:11434/v1` *(veraltet)* | **`…178.103…`** |
+| `AI_BASE_URL` | `http://192.168.178.103:11434/v1` | `…178.103…` |
 | `JETSON_API_KEY` | — | Bearer für Ollama (optional) |
 | `ANTHROPIC_API_KEY` | — | nur bei provider=anthropic |
 | `GITEA_API` | `https://git.eulernest.eu/api/v1` | Bot: intern `http://gitea:3000/api/v1` |

scripts/llm.py

@@ -4,18 +4,18 @@ Bewusst NUR stdlib (urllib) — so braucht der CI-Job kein `pip install` und
 läuft in jedem schlanken Python-Image sofort.
 
 Umschalten zwischen Backends über Env-Vars:
-  AI_PROVIDER=anthropic   -> api.anthropic.com/v1/messages   (Claude, jetzt)
+  AI_PROVIDER=ollama      -> <AI_BASE_URL>/chat/completions   (Jetson, self-hosted, Default)
-  AI_PROVIDER=ollama      -> <AI_BASE_URL>/chat/completions   (Jetson, später)
+  AI_PROVIDER=anthropic   -> api.anthropic.com/v1/messages    (Claude, optional)
 
 Weitere Env-Vars:
-  AI_MODEL       Modellname (Default je nach Provider)
+  AI_MODEL       Modellname (Default je nach Provider; ollama: "coder" = qwen2.5-coder:7b)
-  ANTHROPIC_API_KEY   (für anthropic)
   AI_BASE_URL    OpenAI-kompatible Basis-URL (für ollama),
-                 Default http://192.168.178.20:11434/v1
+                 Default http://192.168.178.103:11434/v1
   JETSON_API_KEY Bearer-Token (für ollama, optional)
+  ANTHROPIC_API_KEY   (für anthropic)
 
-Der Wechsel auf den Jetson ist damit ein reiner Env-Change in der Workflow-
+Provider-Wechsel ist damit ein reiner Env-Change in der Workflow-Datei
-Datei — KEINE Code-Änderung.
+— KEINE Code-Änderung.
 """
 import json
 import os
@@ -29,7 +29,7 @@ def configured():
     Erlaubt den Workflows, den KI-Schritt sauber zu ÜBERSPRINGEN (exit 0) statt
     zu failen, solange z.B. noch kein ANTHROPIC_API_KEY hinterlegt ist.
     """
-    provider = os.environ.get("AI_PROVIDER", "anthropic").lower()
+    provider = os.environ.get("AI_PROVIDER", "ollama").lower()
     if provider == "anthropic":
         if not os.environ.get("ANTHROPIC_API_KEY"):
             return False, "ANTHROPIC_API_KEY ist nicht gesetzt"
@@ -41,12 +41,12 @@ def configured():
 
 def complete(system: str, user: str, *, max_tokens: int = 1500) -> str:
     """Schickt (system, user) ans konfigurierte Backend, gibt reinen Text zurück."""
-    provider = os.environ.get("AI_PROVIDER", "anthropic").lower()
+    provider = os.environ.get("AI_PROVIDER", "ollama").lower()
     if provider == "anthropic":
         model = os.environ.get("AI_MODEL", "claude-3-5-haiku-latest")
         return _anthropic(system, user, model, max_tokens)
     if provider == "ollama":
-        model = os.environ.get("AI_MODEL", "qwen-light")
+        model = os.environ.get("AI_MODEL", "coder")
         return _ollama_openai(system, user, model, max_tokens)
     raise SystemExit(f"[llm] Unbekannter AI_PROVIDER: {provider!r}")
 
@@ -88,7 +88,7 @@ def _anthropic(system: str, user: str, model: str, max_tokens: int) -> str:
 
 
 def _ollama_openai(system: str, user: str, model: str, max_tokens: int) -> str:
-    base = os.environ.get("AI_BASE_URL", "http://192.168.178.20:11434/v1").rstrip("/")
+    base = os.environ.get("AI_BASE_URL", "http://192.168.178.103:11434/v1").rstrip("/")
     headers = {"content-type": "application/json"}
     key = os.environ.get("JETSON_API_KEY", "")
     if key: