feat: legg til direkte Ollama-verktøy for Emma (gemma3:27b), gemma3:4b og qwen2.5:3b på emma-gpu-vm

opax-mcp/server.py

@@ -17,7 +17,7 @@ import logging
 logging.basicConfig(level=logging.INFO)
 logger = logging.getLogger(__name__)
 
-app = FastAPI(title="opax-mcp", version="3.0.2")
+app = FastAPI(title="opax-mcp", version="3.1.0")
 
 OPAX_BASE_URL        = os.environ.get("OPAX_BASE_URL", "https://opax.vauco.no")
 OPAX_IAP_CLIENT_ID   = os.environ.get("OPAX_IAP_CLIENT_ID", "")
@@ -27,7 +27,15 @@ GITEA_TOKEN          = os.environ.get("GITEA_TOKEN", "")
 GITEA_REPO           = os.environ.get("GITEA_REPO", "chris/OSVauco")
 GOOGLE_CLOUD_PROJECT = os.environ.get("GOOGLE_CLOUD_PROJECT", "propane-will-491900-m5")
 
+# Emma-vm Ollama — direkte tilkobling
+OLLAMA_BASE_URL      = os.environ.get("OLLAMA_BASE_URL", "http://34.13.238.133:11434")
+EMMA_MODEL           = os.environ.get("EMMA_MODEL", "gemma3:27b")
+EMMA_FAST_MODEL      = os.environ.get("EMMA_FAST_MODEL", "gemma3:4b")
+EMMA_LIGHT_MODEL     = os.environ.get("EMMA_LIGHT_MODEL", "qwen2.5:3b")
+
 logger.info(f"OPAX_IAP_CLIENT_ID: {OPAX_IAP_CLIENT_ID[:10]}...")
+logger.info(f"OLLAMA_BASE_URL: {OLLAMA_BASE_URL}")
+logger.info(f"EMMA_MODEL: {EMMA_MODEL}")
 
 
 # ---------------------------------------------------------------------------
@@ -37,11 +45,9 @@ logger.info(f"OPAX_IAP_CLIENT_ID: {OPAX_IAP_CLIENT_ID[:10]}...")
 def _verify_auth(request: Request):
     if not MCP_SECRET:
         return
-    # Authorization: Bearer <token>
     auth = request.headers.get("Authorization", "")
     if auth.startswith("Bearer ") and auth[7:] == MCP_SECRET:
         return
-    # X-MCP-Secret or api-key
     if request.headers.get("X-MCP-Secret") == MCP_SECRET:
         return
     if request.headers.get("api-key") == MCP_SECRET:
@@ -88,6 +94,46 @@ async def _opax_post(path: str, body: dict) -> Any:
         return r.json()
 
 
+# ---------------------------------------------------------------------------
+# Ollama helpers — direkte mot emma-gpu-vm
+# ---------------------------------------------------------------------------
+
+async def _ollama_chat(model: str, prompt: str, system: str = "") -> dict:
+    """Kall Ollama chat-API direkte på emma-gpu-vm."""
+    messages = []
+    if system:
+        messages.append({"role": "system", "content": system})
+    messages.append({"role": "user", "content": prompt})
+
+    payload = {
+        "model": model,
+        "messages": messages,
+        "stream": False,
+    }
+    try:
+        async with httpx.AsyncClient(timeout=120) as c:
+            r = await c.post(f"{OLLAMA_BASE_URL}/api/chat", json=payload)
+            r.raise_for_status()
+            data = r.json()
+            return {
+                "model": model,
+                "response": data.get("message", {}).get("content", ""),
+                "done": data.get("done", True),
+                "total_duration_ms": round(data.get("total_duration", 0) / 1e6),
+            }
+    except Exception as e:
+        logger.error(f"[OLLAMA ERROR] model={model} {type(e).__name__}: {e}")
+        raise
+
+
+async def _ollama_models() -> list:
+    """List alle modeller tilgjengelig på emma-gpu-vm."""
+    async with httpx.AsyncClient(timeout=10) as c:
+        r = await c.get(f"{OLLAMA_BASE_URL}/api/tags")
+        r.raise_for_status()
+        return r.json().get("models", [])
+
+
 # ---------------------------------------------------------------------------
 # Gitea helpers
 # ---------------------------------------------------------------------------
@@ -157,7 +203,32 @@ async def run_jason(p):
     return await _opax_post("/run", {"message": p.get("prompt", p.get("message", "")), "user_id": p.get("user_id", "opax"), "session_id": p.get("session_id", "mcp"), "mode": p.get("mode", "light")})
 
 async def run_emma(p):
-    return await _opax_post("/emma", {"message": p.get("prompt", p.get("message", "")), "session_id": p.get("session_id", "mcp")})
+    """Emma Vauger — Gemma 3 27B via Ollama på emma-gpu-vm. Primær lokal AI."""
+    return await _ollama_chat(
+        model=EMMA_MODEL,
+        prompt=p.get("prompt", p.get("message", "")),
+        system=p.get("system", "Du er Emma Vauger, en intelligent og hjelpsom AI-assistent for Vauco. Du er lokal, rask og priveråd.")
+    )
+
+async def run_emma_fast(p):
+    """Emma Fast — Gemma 3 4B via Ollama. Rask inferens for enkle oppgaver."""
+    return await _ollama_chat(
+        model=EMMA_FAST_MODEL,
+        prompt=p.get("prompt", p.get("message", "")),
+        system=p.get("system", "Du er en rask og konsis AI-assistent. Svar kort og presist.")
+    )
+
+async def run_qwen(p):
+    """Qwen 2.5 3B via Ollama. Lett hjelper for enkle spørsmål og kodeoppgaver."""
+    return await _ollama_chat(
+        model=EMMA_LIGHT_MODEL,
+        prompt=p.get("prompt", p.get("message", "")),
+        system=p.get("system", "")
+    )
+
+async def list_emma_models(p):
+    """List alle Ollama-modeller tilgjengelig på emma-gpu-vm."""
+    return await _ollama_models()
 
 async def get_health(p):       return await _opax_get("/health")
 async def get_build_status(p): return await _opax_get("/opax/build-status")
@@ -221,7 +292,13 @@ TOOLS = {
     "send_sms":                    (send_sms,                    "Send SMS", {"type":"object","properties":{"to":{"type":"string"},"message":{"type":"string"}},"required":["to","message"]}),
     "get_notify_channels":         (get_notify_channels,         "Hent varslingkanaler", {}),
     "run_jason":                   (run_jason,                   "Kjør Jason-agenten med en prompt", {"type":"object","properties":{"prompt":{"type":"string"},"mode":{"type":"string"}},"required":["prompt"]}),
-    "run_emma":                    (run_emma,                    "Kjør Emma-agenten med en prompt", {"type":"object","properties":{"prompt":{"type":"string"}},"required":["prompt"]}),
+
+    # — Emma Vauger — direkte Ollama på emma-gpu-vm (34.13.238.133:11434)
+    "run_emma":                    (run_emma,                    "Emma Vauger (gemma3:27b) — primær lokal AI, ingen quota", {"type":"object","properties":{"prompt":{"type":"string"},"system":{"type":"string","description":"Valgfri systemprompt"}},"required":["prompt"]}),
+    "run_emma_fast":               (run_emma_fast,               "Emma Fast (gemma3:4b) — rask versjon for enkle oppgaver", {"type":"object","properties":{"prompt":{"type":"string"},"system":{"type":"string"}},"required":["prompt"]}),
+    "run_qwen":                    (run_qwen,                    "Qwen 2.5 3B — lett hjelper og kodeoppgaver", {"type":"object","properties":{"prompt":{"type":"string"},"system":{"type":"string"}},"required":["prompt"]}),
+    "list_emma_models":            (list_emma_models,            "List alle Ollama-modeller tilgjengelig på emma-gpu-vm", {}),
+
     "get_health":                  (get_health,                  "Hent helsestatus for OPAX", {}),
     "get_build_status":            (get_build_status,            "Hent siste build-status", {}),
     "get_state":                   (get_state,                   "Hent platform-tilstand", {}),
@@ -271,19 +348,16 @@ async def mcp_handler(request: Request):
     req_id = body.get("id")
     params = body.get("params", {})
 
-    # initialize
     if method == "initialize":
         return JSONResponse(_jsonrpc_ok(req_id, {
             "protocolVersion": "2024-11-05",
             "capabilities": {"tools": {}},
-            "serverInfo": {"name": "opax-mcp", "version": "3.0.0"},
+            "serverInfo": {"name": "opax-mcp", "version": "3.1.0"},
         }))
 
-    # tools/list
     if method == "tools/list":
         return JSONResponse(_jsonrpc_ok(req_id, {"tools": _tool_list_result()}))
 
-    # tools/call
     if method == "tools/call":
         tool_name = params.get("name") or params.get("tool")
         tool_args = params.get("arguments", params.get("params", {}))
@@ -303,7 +377,6 @@ async def mcp_handler(request: Request):
             "content": [{"type": "text", "text": json.dumps(result, ensure_ascii=False)}]
         }))
 
-    # notifications (fire-and-forget, no response needed)
     if method.startswith("notifications/"):
         return JSONResponse(status_code=202, content={})
 
@@ -316,4 +389,4 @@ async def mcp_handler(request: Request):
 
 @app.get("/health")
 async def health():
-    return {"status": "ok", "service": "opax-mcp", "version": "3.0.0"}
+    return {"status": "ok", "service": "opax-mcp", "version": "3.1.0", "ollama": OLLAMA_BASE_URL, "emma_model": EMMA_MODEL}