add links for PDFs

app.py

@@ -262,17 +262,20 @@ Answer:"""
                         )
                         if meta.get("type") == "url" and meta.get("source"):
                             context_index_to_url[i] = meta["source"]
+                        elif meta.get("url"):
+                            context_index_to_url[i] = meta["url"]
 
                     # Generate response using chat/comversational API (Mistral instruct uses this)
                     try:
                         response_text = self._generate_with_chat(prompt, max_new_tokens=512)
                         if response_text:
                             # Resolve [Context N] to actual source labels in the body
-                            # First, expand grouped "Context N, M" forms (without brackets) to individual [Context N] tags
+                            # Expand bare/grouped "Context N, M" forms to individual [Context N] tags.
+                            # Use negative lookbehind (?<!\[) so already-bracketed [Context N] are not double-wrapped.
                             def _expand_grouped_context(m):
                                 indices = [int(x.strip()) for x in re.split(r'[,\s]+', m.group(1)) if x.strip().isdigit()]
                                 return " ".join(f"[Context {idx}]" for idx in indices)
-                            response_text = re.sub(r'\bContext\s+([\d][,\s\d]*)', _expand_grouped_context, response_text)
+                            response_text = re.sub(r'(?<!\[)\bContext\s+([\d][,\s\d]*)', _expand_grouped_context, response_text)
                             # Now replace all [Context N] with source labels/links
                             for i, source_label in context_index_to_source.items():
                                 url = context_index_to_url.get(i)

data/vector_store/documents.pkl

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:91f5564b211362f8521a1f7941a4f346d97d46dfba2e83d8ca58cb81a77df864
-size 874711
+oid sha256:0ba58f4cd0cf25aec81a1c328e116e6276dd5648819412e80a9d9cbf9e9a109a
+size 879299

data/vector_store/embeddings.pkl

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:a043f7bc5e7b5f5a2880cc6d5389fb277f9529eac26612880828b241970d2b9f
+oid sha256:d6ad8b4a984849be946ed42dc015bc669d4882712e89318f72ea65e4fbc1d79a
 size 4390028

data/vector_store/index.faiss

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:73c8f7fdb095c972962c80cc05f36a80bec95d374736c8e762f165621185d9f8
+oid sha256:06988afb211c9b69be9e929e45c1b13dbf48cb0b9549b30b24e9f15d78aadcbb
 size 4389933

ingest_documents.py

@@ -17,6 +17,13 @@ URLS = [
     "https://cog.ca/faqs/"
 ]
 
+# Optional: map PDF filenames to their publicly hosted URLs so references are hyperlinked.
+# Keys are bare filenames (no path), values are the public URL for that PDF.
+PDF_URLS = {
+    "Organic production systems - General principles and management standards.pdf": "https://publications.gc.ca/collections/collection_2026/ongc-cgsb/P29-32-310-2026-eng.pdf",
+    # "another-doc.pdf": "https://example.com/another-doc.pdf",
+}
+
 
 def main():
     """Main ingestion function."""
@@ -61,7 +68,7 @@ def main():
     print("=" * 60)
     
     try:
-        documents = ingestion.process_documents(pdf_paths=pdf_paths, urls=urls)
+        documents = ingestion.process_documents(pdf_paths=pdf_paths, urls=urls, pdf_urls=PDF_URLS)
         print(f"\n[SUCCESS] Successfully processed {len(documents)} document chunks")
         
         # Build vector store

ingestion.py

@@ -115,20 +115,22 @@ class DocumentIngestion:
         except Exception as e:
             raise Exception(f"Error reading URL {url}: {str(e)}")
     
-    def process_documents(self, pdf_paths: List[str] = None, urls: List[str] = None) -> List[Dict]:
+    def process_documents(self, pdf_paths: List[str] = None, urls: List[str] = None, pdf_urls: Dict[str, str] = None) -> List[Dict]:
         """
         Process PDFs and URLs into chunks.
-        
+
         Args:
             pdf_paths: List of PDF file paths
             urls: List of URLs to process
-            
+            pdf_urls: Optional dict mapping PDF filenames to their public URLs (for hyperlinking references)
+
         Returns:
             List of document chunks with metadata
         """
         all_texts = []
         all_metadata = []
-        
+        pdf_urls = pdf_urls or {}
+
         # Process PDFs
         if pdf_paths:
             for pdf_path in pdf_paths:
@@ -138,14 +140,19 @@ class DocumentIngestion:
                 document_title = self.get_pdf_document_title(pdf_path)
                 text = self.read_pdf(pdf_path)
                 chunks = self.text_splitter.split_text(text)
+                filename = os.path.basename(pdf_path)
+                public_url = pdf_urls.get(filename)
                 for i, chunk in enumerate(chunks):
                     all_texts.append(chunk)
-                    all_metadata.append({
+                    meta = {
                         'source': pdf_path,
                         'document_title': document_title,
                         'type': 'pdf',
                         'chunk_index': i
-                    })
+                    }
+                    if public_url:
+                        meta['url'] = public_url
+                    all_metadata.append(meta)
         
         # Process URLs
         if urls: