Update src/ingestion.py
Browse files- src/ingestion.py +19 -5
src/ingestion.py
CHANGED
|
@@ -131,19 +131,20 @@ from gen_ai_hub.proxy.langchain.openai import ChatOpenAI
|
|
| 131 |
def adaptive_fallback_toc(text: str, model_name: str = "gpt-4o"):
|
| 132 |
"""
|
| 133 |
Uses SAP GenAI Hub proxy (same as QA pipeline) to infer a Table of Contents.
|
| 134 |
-
This
|
|
|
|
| 135 |
"""
|
| 136 |
snippet = text[:7000]
|
| 137 |
|
| 138 |
-
|
| 139 |
-
creds_path = os.path.join(os.path.dirname(__file__), "GEN AI HUB PROXY.json")
|
| 140 |
base_url = ""
|
| 141 |
|
|
|
|
|
|
|
| 142 |
if os.path.exists(creds_path):
|
| 143 |
try:
|
| 144 |
with open(creds_path, "r") as f:
|
| 145 |
creds = json.load(f)
|
| 146 |
-
# Try all known locations for base URL
|
| 147 |
base_url = (
|
| 148 |
creds.get("base_url")
|
| 149 |
or creds.get("serviceurls", {}).get("AI_API_URL", "")
|
|
@@ -151,11 +152,23 @@ def adaptive_fallback_toc(text: str, model_name: str = "gpt-4o"):
|
|
| 151 |
)
|
| 152 |
except Exception as e:
|
| 153 |
print(f"⚠️ Could not read GenAI proxy credentials: {e}")
|
|
|
|
|
|
|
|
|
|
| 154 |
|
| 155 |
if not base_url:
|
| 156 |
print("⚠️ Missing AI_API_URL or base_url in credentials — skipping fallback.")
|
| 157 |
return []
|
| 158 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 159 |
try:
|
| 160 |
print(f"⚙️ Invoking GenAI proxy for TOC inference using model: {model_name}")
|
| 161 |
proxy_client = get_proxy_client("gen-ai-hub", base_url=base_url)
|
|
@@ -179,7 +192,7 @@ def adaptive_fallback_toc(text: str, model_name: str = "gpt-4o"):
|
|
| 179 |
response = llm.invoke(prompt)
|
| 180 |
response_text = getattr(response, "content", str(response))
|
| 181 |
|
| 182 |
-
# Extract clean TOC-like lines
|
| 183 |
lines = [
|
| 184 |
re.sub(r"^[0-9.\-•\s]+", "", l.strip())
|
| 185 |
for l in response_text.splitlines()
|
|
@@ -195,6 +208,7 @@ def adaptive_fallback_toc(text: str, model_name: str = "gpt-4o"):
|
|
| 195 |
return []
|
| 196 |
|
| 197 |
|
|
|
|
| 198 |
# ==========================================================
|
| 199 |
# 3B️⃣ UNIFIED WRAPPER (Heuristic + AI Hybrid)
|
| 200 |
# ==========================================================
|
|
|
|
| 131 |
def adaptive_fallback_toc(text: str, model_name: str = "gpt-4o"):
|
| 132 |
"""
|
| 133 |
Uses SAP GenAI Hub proxy (same as QA pipeline) to infer a Table of Contents.
|
| 134 |
+
This ensures consistent credentials, no manual token handling, and safe reuse
|
| 135 |
+
of your existing GEN AI HUB PROXY.json configuration.
|
| 136 |
"""
|
| 137 |
snippet = text[:7000]
|
| 138 |
|
| 139 |
+
creds = {}
|
|
|
|
| 140 |
base_url = ""
|
| 141 |
|
| 142 |
+
# ✅ Load credentials from same JSON as QA pipeline
|
| 143 |
+
creds_path = os.path.join(os.path.dirname(__file__), "GEN AI HUB PROXY.json")
|
| 144 |
if os.path.exists(creds_path):
|
| 145 |
try:
|
| 146 |
with open(creds_path, "r") as f:
|
| 147 |
creds = json.load(f)
|
|
|
|
| 148 |
base_url = (
|
| 149 |
creds.get("base_url")
|
| 150 |
or creds.get("serviceurls", {}).get("AI_API_URL", "")
|
|
|
|
| 152 |
)
|
| 153 |
except Exception as e:
|
| 154 |
print(f"⚠️ Could not read GenAI proxy credentials: {e}")
|
| 155 |
+
else:
|
| 156 |
+
print("⚠️ No SAP GenAI credentials file found — skipping AI fallback.")
|
| 157 |
+
return []
|
| 158 |
|
| 159 |
if not base_url:
|
| 160 |
print("⚠️ Missing AI_API_URL or base_url in credentials — skipping fallback.")
|
| 161 |
return []
|
| 162 |
|
| 163 |
+
# ✅ Inject credentials into environment (matches QA setup)
|
| 164 |
+
os.environ.update({
|
| 165 |
+
"AICORE_AUTH_URL": creds.get("url", ""),
|
| 166 |
+
"AICORE_CLIENT_ID": creds.get("clientid") or creds.get("client_id", ""),
|
| 167 |
+
"AICORE_CLIENT_SECRET": creds.get("clientsecret") or creds.get("client_secret", ""),
|
| 168 |
+
"AICORE_RESOURCE_GROUP": "default",
|
| 169 |
+
"AICORE_BASE_URL": base_url
|
| 170 |
+
})
|
| 171 |
+
|
| 172 |
try:
|
| 173 |
print(f"⚙️ Invoking GenAI proxy for TOC inference using model: {model_name}")
|
| 174 |
proxy_client = get_proxy_client("gen-ai-hub", base_url=base_url)
|
|
|
|
| 192 |
response = llm.invoke(prompt)
|
| 193 |
response_text = getattr(response, "content", str(response))
|
| 194 |
|
| 195 |
+
# ✅ Extract clean TOC-like lines
|
| 196 |
lines = [
|
| 197 |
re.sub(r"^[0-9.\-•\s]+", "", l.strip())
|
| 198 |
for l in response_text.splitlines()
|
|
|
|
| 208 |
return []
|
| 209 |
|
| 210 |
|
| 211 |
+
|
| 212 |
# ==========================================================
|
| 213 |
# 3B️⃣ UNIFIED WRAPPER (Heuristic + AI Hybrid)
|
| 214 |
# ==========================================================
|