Merge pull request #9 from webstackdev/feature/embeddings

Feature/embeddings

Author: webstackdev

.env.example

@@ -11,6 +11,16 @@ OPENROUTER_API_KEY=
 OPENROUTER_API_BASE=https://openrouter.ai/api/v1
 OPENROUTER_APP_URL=
 OPENROUTER_APP_NAME=newsletter-maker
+AI_CLASSIFICATION_MODEL=meta-llama/llama-3.1-70b-instruct
+AI_RELEVANCE_MODEL=qwen/qwen-2.5-72b-instruct
+AI_SUMMARIZATION_MODEL=google/gemma-3-27b-it
+AI_CLASSIFICATION_REVIEW_THRESHOLD=0.6
+AI_RELEVANCE_LOW_THRESHOLD=0.5
+AI_RELEVANCE_HIGH_THRESHOLD=0.85
+AI_RELEVANCE_REVIEW_THRESHOLD=0.4
+AI_RELEVANCE_SUMMARIZE_THRESHOLD=0.7
+AI_MAX_NODE_RETRIES=2
+AI_REQUEST_TIMEOUT_SECONDS=60
 EMBEDDING_PROVIDER=sentence-transformers
 EMBEDDING_MODEL=sentence-transformers/all-MiniLM-L6-v2
 EMBEDDING_TRUST_REMOTE_CODE=false

.vscode/settings.json

@@ -18,6 +18,7 @@
     "noinput",
     "nomic",
     "OLLAMA",
+    "ormsgpack",
     "PRAW",
     "psycopg",
     "pylint",
@@ -28,6 +29,8 @@
     "readyz",
     "Referer",
     "upserted",
-    "upvote"
+    "upvote",
+    "uritemplate",
+    "xxhash"
   ]
 }

celerybeat-schedule

@@ -0,0 +1,71 @@
+from __future__ import annotations
+
+import json
+import re
+import time
+from dataclasses import dataclass
+from typing import Any
+
+import httpx
+from django.conf import settings
+
+JSON_OBJECT_PATTERN = re.compile(r"\{.*\}", re.DOTALL)
+
+
+@dataclass(slots=True)
+class OpenRouterJSONResponse:
+    payload: dict[str, Any]
+    model: str
+    latency_ms: int
+
+
+def openrouter_chat_json(*, model: str, system_prompt: str, user_prompt: str) -> OpenRouterJSONResponse:
+    if not settings.OPENROUTER_API_KEY:
+        raise RuntimeError("OPENROUTER_API_KEY must be configured for OpenRouter chat completions.")
+
+    headers = {
+        "Authorization": f"Bearer {settings.OPENROUTER_API_KEY}",
+        "Content-Type": "application/json",
+    }
+    if settings.OPENROUTER_APP_URL:
+        headers["HTTP-Referer"] = settings.OPENROUTER_APP_URL
+    if settings.OPENROUTER_APP_NAME:
+        headers["X-OpenRouter-Title"] = settings.OPENROUTER_APP_NAME
+
+    started_at = time.perf_counter()
+    response = httpx.post(
+        f"{settings.OPENROUTER_API_BASE.rstrip('/')}/chat/completions",
+        headers=headers,
+        json={
+            "model": model,
+            "temperature": 0,
+            "response_format": {"type": "json_object"},
+            "messages": [
+                {"role": "system", "content": system_prompt},
+                {"role": "user", "content": user_prompt},
+            ],
+        },
+        timeout=settings.AI_REQUEST_TIMEOUT_SECONDS,
+    )
+    latency_ms = int((time.perf_counter() - started_at) * 1000)
+    response.raise_for_status()
+
+    message_content = response.json()["choices"][0]["message"]["content"]
+    return OpenRouterJSONResponse(
+        payload=_extract_json_object(message_content),
+        model=model,
+        latency_ms=latency_ms,
+    )
+
+
+def _extract_json_object(message_content: str) -> dict[str, Any]:
+    try:
+        payload = json.loads(message_content)
+    except json.JSONDecodeError:
+        match = JSON_OBJECT_PATTERN.search(message_content)
+        if not match:
+            raise ValueError("Model response did not contain a JSON object.")
+        payload = json.loads(match.group(0))
+    if not isinstance(payload, dict):
+        raise ValueError("Model response JSON must be an object.")
+    return payload