File size: 4,853 Bytes
c4e6d00 | 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 | import pandas as pd
import json
import os
# On définit les chemins par rapport à la racine du projet
ROOT_DIR = os.path.dirname(os.path.dirname(os.path.abspath(__file__)))
INPUT_DIR = os.path.join(ROOT_DIR, "data", "input")
OUTPUT_JSONL = os.path.join(ROOT_DIR, "data", "metiers_rag.jsonl")
OUTPUT_PARQUET = os.path.join(ROOT_DIR, "data", "metiers_rag.parquet")
def prepare_expert_rag_jsonl(max_words=300):
print(f"🚀 Démarrage de la préparation (Racine: {ROOT_DIR})")
try:
df_metiers = pd.read_csv(os.path.join(INPUT_DIR, "gem_metiers.csv"))
df_liens = pd.read_csv(os.path.join(INPUT_DIR, "gem_metiers_competences.csv"))
df_dict_comp = pd.read_csv(os.path.join(INPUT_DIR, "gem_competences.csv"))
df_verbatims = pd.DataFrame(columns=["id_metier", "verbatim"])
v_path = os.path.join(INPUT_DIR, "metiers_verbatims.csv")
if os.path.exists(v_path):
df_verbatims = pd.read_csv(v_path)
print(f"✅ {len(df_verbatims)} verbatims trouvés.")
except Exception as e:
print(f"❌ Erreur : {e}")
return
comp_defs = {}
for _, row in df_dict_comp.iterrows():
c_id = row['Id interne Neobrain Compétence']
comp_defs[c_id] = {i: row.get(f'Description Niveau {i} Compétence', '') for i in range(1, 5)}
rag_items = []
for _, metier in df_metiers.iterrows():
code_m = metier['Code Métier']
nom_m = metier['Métier collaborateur']
famille = metier['Famille métier']
comp_du_metier = df_liens[df_liens['Code Métier'] == code_m].sort_values(by='Poids de la compétence', ascending=False)
header = f"# Fiche Métier : {nom_m}\n\n"
header += f"- **Code Métier** : {code_m}\n"
header += f"- **Famille** : {famille}\n\n"
v_list = df_verbatims[df_verbatims['id_metier'] == code_m]['verbatim'].tolist()
if v_list:
header += "## Missions quotidiennes (Verbatims)\n"
for v in v_list:
header += f"- \"{v}\"\n"
header += "\n"
header += "## Profil de compétences (Attentes)\n\n"
comp_lines = []
for groupe, df_g in comp_du_metier.groupby('Groupe de compétences', sort=False):
comp_lines.append(f"### {groupe}")
for _, row_c in df_g.iterrows():
nom_c = row_c['Nom de la compétence']
val = row_c['Niveau requis pour le métier']
niv_req = int(float(val)) if pd.notna(val) else 0
c_id = row_c['Id interne Neobrain Compétence']
desc_niveau = comp_defs.get(c_id, {}).get(niv_req, "")
line = f"- **{nom_c}** (Niveau {niv_req})"
if pd.notna(desc_niveau) and desc_niveau != "":
line += f" : {desc_niveau}"
comp_lines.append(line)
comp_lines.append("")
current_chunk_comps = []
current_word_count = len(header.split())
chunk_id = 1
def add_item(cid, text, midx):
rag_items.append({
"id": f"{cid}_{midx}" if midx > 1 else cid,
"title": nom_m,
"text": text,
"metadata": {"famille": famille, "chunk": midx, "source": "OPT-NC Expert Ref 2025"}
})
for line in comp_lines:
line_words = len(line.split())
if current_word_count + line_words > max_words and current_chunk_comps:
add_item(code_m, header + "\n".join(current_chunk_comps), chunk_id)
chunk_id += 1
current_chunk_comps = [line]
current_word_count = len(header.split()) + line_words
else:
current_chunk_comps.append(line)
current_word_count += line_words
if current_chunk_comps:
add_item(code_m, header + "\n".join(current_chunk_comps), chunk_id)
# EXPORT DOUBLE FORMAT
os.makedirs(os.path.dirname(OUTPUT_JSONL), exist_ok=True)
# 1. JSONL
with open(OUTPUT_JSONL, "w", encoding="utf-8") as f:
for item in rag_items:
f.write(json.dumps(item, ensure_ascii=False) + "\n")
# 2. PARQUET (On aplatit un peu pour le format colonnaire)
df_final = pd.DataFrame(rag_items)
df_final['famille'] = df_final['metadata'].apply(lambda x: x['famille'])
df_final['source'] = df_final['metadata'].apply(lambda x: x['source'])
df_final['chunk_index'] = df_final['metadata'].apply(lambda x: x['chunk'])
df_final.drop(columns=['metadata']).to_parquet(OUTPUT_PARQUET, index=False)
print(f"✨ Terminé !")
print(f"📄 JSONL : {OUTPUT_JSONL}")
print(f"📦 PARQUET : {OUTPUT_PARQUET}")
if __name__ == "__main__":
prepare_expert_rag_jsonl()
|