Files
agentic/app/core/embedding_engine.py
T
2026-06-10 08:20:27 +02:00

80 lines
2.9 KiB
Python

"""Embedding generation via Ollama's qwen3-embedding:8b model."""
import hashlib
import numpy as np
def get_embedding_sync(text: str, ollama_url: str = "http://10.0.1.127:11434") -> list[float]:
"""Synchronous embedding call via Ollama /qwen3-embedding:8b."""
try:
import httpx
with httpx.Client(timeout=30) as client:
resp = client.post(
f"{ollama_url}/api/embed",
json={"model": "qwen3-embedding:8b", "input": text},
)
resp.raise_for_status()
data = resp.json()
vectors = data.get("embeddings", [])
if vectors:
# Ollama may return multiple inputs; use first
emb = vectors[0] if isinstance(vectors[0], list) else vectors
return emb
except Exception:
pass
# Fallback to deterministic feature vector if Ollama unavailable
return _hash_embedding(text)
def get_embedding(text: str, ollama_url: str = "http://10.0.1.127:11434") -> list[float]:
"""Generate an embedding — calls Ollama qwen3-embedding:8b."""
return get_embedding_sync(text, ollama_url)
async def get_embedding_async(text: str, ollama_url: str = "http://10.0.1.127:11434") -> list[float]:
"""Async embedding call via Ollama /qwen3-embedding:8b."""
try:
import httpx
async with httpx.AsyncClient(timeout=30) as client:
resp = await client.post(
f"{ollama_url}/api/embed",
json={"model": "qwen3-embedding:8b", "input": text},
)
resp.raise_for_status()
data = resp.json()
vectors = data.get("embeddings", [])
if vectors:
emb = vectors[0] if isinstance(vectors[0], list) else vectors
return emb
except Exception:
pass
return _hash_embedding(text)
def _hash_embedding(text: str) -> list[float]:
"""Deterministic 4096-dim feature vector fallback (no Ollama needed)."""
feature_dim = 4096
vec = np.zeros(feature_dim, dtype=np.float32)
for n in [1, 2, 3, 4]:
tokens = [text[i:i+n] for i in range(len(text)-n+1)]
for token in tokens[:200]:
h = hashlib.md5(token.encode()).hexdigest()
for i in range(0, 12, 3):
val = (int(h[i:i+2], 16) - 128) / 128.0
feature_idx = (int(h[i+2:i+4], 16) * 37) % feature_dim
vec[feature_idx] += val
norm = np.linalg.norm(vec)
if norm > 0:
vec /= norm
return vec.tolist()
def compute_similarity(vec1: list[float], vec2: list[float]) -> float:
"""Cosine similarity between two vectors."""
v1 = np.array(vec1, dtype=np.float32)
v2 = np.array(vec2, dtype=np.float32)
if v1.shape[0] != v2.shape[0]:
return 0.0
v1 /= np.linalg.norm(v1) + 1e-8
v2 /= np.linalg.norm(v2) + 1e-8
return float(np.dot(v1, v2))