File size: 8,350 Bytes
9df97a2 | 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 | import json
from fastapi import APIRouter, HTTPException, UploadFile, File, Form
from typing import Dict, Any, Optional
from app.services.cv_extractor import CVExtractionService
from app.services.feature_engineering import PairFeatureMeta, build_pair_features, fit_pair_vectorizer
from app.services.matching_service import MatchingService
from app.services.scoring import compute_match_score, apply_business_rules
from app.scoring.decision import combine_scores, decision_from_score
import joblib
import pickle
from pathlib import Path
import numpy as _np
router = APIRouter(prefix="/api/pipeline", tags=["pipeline"])
def _load_bert_first_model_bundle():
"""Load the BERT-first pairwise scoring bundle if it exists."""
models_dir = Path("models/classical_from_tfidf_bertfirst")
if not models_dir.exists():
return None
meta_path = models_dir / "pair_feature_meta.joblib"
if not meta_path.exists():
return None
try:
meta = joblib.load(meta_path)
except Exception:
return None
model = None
for filename in ("xgboost.model", "random_forest.joblib", "logistic.joblib"):
model_path = models_dir / filename
if model_path.exists():
try:
model = joblib.load(model_path)
except Exception:
try:
import xgboost as xgb
model = xgb.Booster()
model.load_model(str(model_path))
except Exception:
model = None
if model is not None:
break
if model is None:
return None
return {"model": model, "meta": meta}
@router.post("/run")
def run_pipeline(payload: Dict[str, Any]):
"""Run full pipeline: extraction -> features -> matching -> scoring -> decision.
Accepts either `candidate` (raw_text or fields) or `candidate_id` (not implemented),
and `job` (job_text + metadata).
"""
candidate = payload.get("candidate")
job = payload.get("job")
mode = payload.get("mode", "semantic")
if not candidate or not job:
raise HTTPException(status_code=400, detail="Both 'candidate' and 'job' are required in payload")
# Extraction (if raw text provided)
extractor = CVExtractionService()
if isinstance(candidate, dict) and candidate.get("raw_text"):
extraction = extractor.extract_from_text(candidate.get("raw_text"))
else:
# If structured candidate provided, build a lightweight extraction result
extraction = extractor.extract_from_text(candidate.get("raw_text", ""))
cv_text = extraction.raw_text or ""
job_text = job.get("job_text") or job.get("description") or ""
# Feature engineering: fit a small TF-IDF meta on-the-fly when needed
meta = fit_pair_vectorizer([cv_text], [job_text])
features = build_pair_features(cv_text, job_text, meta)
# Matching
matcher = MatchingService()
if mode == "semantic":
sim = matcher.semantic_similarity(cv_text, job_text)
elif mode == "vector":
# Keep the vector mode aligned with the BERT-first matching stack.
sim = matcher.semantic_similarity(cv_text, job_text)
elif mode == "continuous":
cont = matcher.continuous_similarity(job_text, cv_text)
sim = cont.get("max", cont.get("mean", 0.0))
else:
sim = matcher.deep_match_score(job_text, cv_text)
# Scoring and decision
cv_skills = extraction.structured.get("skills", []) if extraction.structured else []
job_skills = job.get("skills", [])
cv_years = extraction.structured.get("years_experience", 0) if extraction.structured else 0
job_years = job.get("years_experience", 0)
score = compute_match_score(
cv_skills=cv_skills,
job_skills=job_skills,
cv_years=cv_years,
job_years=job_years,
similarity_score=float(sim),
)
# Try to augment with ML model score (if trained models present)
ml_score_pct = None
model_bundle = _load_bert_first_model_bundle()
if model_bundle is not None:
try:
clf = model_bundle["model"]
meta = model_bundle["meta"]
X = build_pair_features(cv_text, job_text, meta)
try:
prob = clf.predict_proba(X)[:, 1].ravel()[0]
ml_score_pct = float(prob * 100.0)
except Exception:
try:
pred = clf.predict(X).ravel()[0]
ml_score_pct = float(pred * 100.0)
except Exception:
ml_score_pct = None
except Exception:
ml_score_pct = None
# ml_score_pct None -> map from heuristic of extraction quality
if ml_score_pct is None:
ml_score_pct = float(extraction.quality_score or 0)
# combine similarity (0..1) -> convert to 0..100
sim_pct = float(sim) * 100.0 if sim is not None else 0.0
final_score = combine_scores(sim_pct, ml_score_pct, w_sim=0.5, w_ml=0.5)
decision_label, decision_meta = decision_from_score(final_score)
return {
"extraction": {
"quality_score": extraction.quality_score,
"structured": extraction.structured,
},
"similarity": float(sim),
"ml_score": ml_score_pct,
"final_score": final_score,
"decision": {"label": decision_label, "meta": decision_meta},
}
def np_dot_cosine(a, b):
import numpy as _np
na = _np.linalg.norm(a)
nb = _np.linalg.norm(b)
if na == 0 or nb == 0:
return 0.0
return float(_np.dot(a, b) / (na * nb))
@router.post("/run-full")
async def run_full(cv: UploadFile = File(...), job_json: str = Form(...)):
"""Minimal endpoint: upload a CV file and a job JSON, return pipeline decision.
This endpoint saves the uploaded CV to a temporary path, runs the existing
`run_pipeline` function by providing extracted raw text, and returns the same
output shape as `/run`.
"""
try:
job = json.loads(job_json)
except Exception:
raise HTTPException(status_code=400, detail="'job_json' must be valid JSON")
if not isinstance(job, dict):
raise HTTPException(status_code=400, detail="'job_json' must decode to an object")
from uuid import uuid4
tmp_path = Path('/tmp')
tmp_path.mkdir(parents=True, exist_ok=True)
suffix = Path(cv.filename).suffix or '.pdf'
dest = tmp_path / f"uploaded_cv_{uuid4().hex}{suffix}"
content = await cv.read()
try:
dest.write_bytes(content)
except Exception as exc:
raise HTTPException(status_code=500, detail=f"Failed to save uploaded file: {exc}")
extractor = CVExtractionService()
try:
extraction = extractor.extract_from_pdf(str(dest))
except Exception:
# fallback: try text extraction
text = content.decode('utf-8', errors='ignore')
extraction = extractor.extract_from_text(text)
payload = {"candidate": {"raw_text": extraction.raw_text or ""}, "job": job}
decision_payload = run_pipeline(payload)
matcher = MatchingService()
job_text = job.get("job_text") or job.get("description") or ""
top_k = int(job.get("top_k", 5))
top_k_results = matcher.search_top_k_candidates(job_text=job_text, top_k=top_k)
return {
"decision": decision_payload,
"top_k": {
"job_text": job_text,
"top_k": top_k,
"results": top_k_results,
},
}
@router.post("/top-k")
def top_k_candidates(payload: Dict[str, Any]):
"""Return the top-K CV candidates for a job description using FAISS.
Expected payload:
{
"job_text": "...",
"top_k": 5,
"index_dir": "models/faiss_index"
}
"""
job_text = payload.get("job_text") or payload.get("description") or ""
top_k = int(payload.get("top_k", 5))
index_dir = payload.get("index_dir", "models/faiss_index")
if not job_text.strip():
raise HTTPException(status_code=400, detail="'job_text' is required")
matcher = MatchingService()
results = matcher.search_top_k_candidates(job_text=job_text, top_k=top_k, index_dir=index_dir)
return {
"job_text": job_text,
"top_k": top_k,
"index_dir": index_dir,
"results": results,
}
|