first commit

Dockerfile

@@ -0,0 +1,10 @@
+FROM langchain/langchain:latest
+
+WORKDIR /app
+COPY requirements.txt .
+RUN pip install --no-cache-dir -r requirements.txt
+
+COPY . .
+
+CMD ["uvicorn", "entrypoint:app", "--host", "0.0.0.0", "--port", "8000"]
+

captain-definition

@@ -0,0 +1,5 @@
+{
+  "schemaVersion": 2,
+  "dockerfilePath": "./Dockerfile"
+}
+

entrypoint.py

@@ -0,0 +1,85 @@
+# main.py
+from qdrant_client.http import models as qmodels
+import uuid
+from fastapi import FastAPI
+import requests, os, redis
+from qdrant_client import QdrantClient
+from langchain_community.vectorstores import Qdrant
+from langchain_community.embeddings import HuggingFaceEmbeddings
+
+app = FastAPI()
+
+# Connectors
+qdrant = QdrantClient(url=os.getenv("QDRANT_URL"))
+r = redis.Redis.from_url(os.getenv("REDIS_URL"))
+SEARXNG_URL = os.getenv("SEARXNG_URL")
+LLAMA_SERVER_URL = os.getenv("LLAMA_SERVER_URL")
+QDRANT_URL = os.getenv("QDRANT_URL")
+REDIS_URL = os.getenv("REDIS_URL")
+
+embeddings = HuggingFaceEmbeddings(model_name="sentence-transformers/all-MiniLM-L6-v2")
+
+@app.get("/ask")
+def ask(query: str):
+    # 1. Search via SearxNG
+    resp = requests.get(f"{SEARXNG_URL}/search", params={"q": query, "format": "json"})
+    snippets = [r["title"] + " " + r["content"] for r in resp.json()["results"][:3]]
+
+    context = " ".join(snippets)
+
+    # 2. Embed + store in Qdrant
+    vectors = embeddings.embed_documents(snippets)
+    points = [
+        qmodels.PointStruct(
+            id=str(uuid.uuid4()),          # unique ID for each snippet
+            vector=vec,                    # the embedding vector
+            payload={"text": text}         # optional metadata
+        )
+        for text, vec in zip(snippets, vectors)
+    ]
+    qdrant.upsert(collection_name="docs", points=points)
+
+    # 3. Call llama-server
+    llama_http = requests.post(
+        f"{LLAMA_SERVER_URL}/completion",
+        json={
+            "prompt": f"Context:\n{context}\n\nQuestion: {query}\nAnswer:",
+            "n_predict": 128,
+            "temperature": 0.7,
+            "stop": ["</s>"]
+        }
+    )
+    llm_resp = llama_http.json()
+    print("Llama raw response:", llm_resp)
+
+    return {"answer": llm_resp.get("content", "")}
+@app.get("/health")
+def health():
+    results = {}
+
+    # Check Redis
+    try:
+        r = redis.Redis.from_url(REDIS_URL)
+        r.ping()
+        results["redis"] = "ok"
+    except Exception as e:
+        results["redis"] = f"error: {e}"
+
+    # Check Qdrant
+    try:
+        q = requests.get(f"{QDRANT_URL}/readyz")
+        results["qdrant"] = f"ok ({q.status_code})"
+    except Exception as e:
+        results["qdrant"] = f"error: {e}"
+
+    # Check SearxNG
+    try:
+        s = requests.get(f"{SEARXNG_URL}/search",
+                         params={"q": "ping", "format": "json"},
+                         headers={"X-Forwarded-For": "127.0.0.1"})
+        results["searxng"] = f"ok ({s.status_code})"
+    except Exception as e:
+        results["searxng"] = f"error: {e}"
+
+    return results
+

requirements.txt

@@ -0,0 +1,8 @@
+fastapi
+uvicorn
+qdrant-client
+redis
+requests
+sentence-transformers
+langchain-community
+