signalwire-agents 0.1.47__py3-none-any.whl → 0.1.49__py3-none-any.whl

signalwire_agents/core/mixins/ai_config_mixin.py

@@ -380,6 +380,7 @@ class AIConfigMixin:
         The server will validate and apply parameters based on the target model's capabilities.
         
         Common parameters include:
+            model: The AI model to use (gpt-4o-mini, gpt-4.1-mini, gpt-4.1-nano, nova-micro, nova-lite)
             temperature: Randomness setting. Lower values make output more deterministic.
             top_p: Alternative to temperature. Controls nucleus sampling.
             barge_confidence: ASR confidence to interrupt. Higher values make it harder to interrupt.
@@ -394,6 +395,7 @@ class AIConfigMixin:
             
         Example:
             agent.set_prompt_llm_params(
+                model="nova-micro",  # Using Amazon's nova-micro model
                 temperature=0.7,
                 top_p=0.9,
                 barge_confidence=0.6
@@ -413,6 +415,7 @@ class AIConfigMixin:
         The server will validate and apply parameters based on the target model's capabilities.
         
         Common parameters include:
+            model: The AI model to use (gpt-4o-mini, gpt-4.1-mini, gpt-4.1-nano, nova-micro, nova-lite)
             temperature: Randomness setting. Lower values make output more deterministic.
             top_p: Alternative to temperature. Controls nucleus sampling.
             presence_penalty: Topic diversity. Positive values encourage new topics.
@@ -427,6 +430,7 @@ class AIConfigMixin:
             
         Example:
             agent.set_post_prompt_llm_params(
+                model="gpt-4o-mini",
                 temperature=0.5,  # More deterministic for post-prompt
                 top_p=0.9
             )

signalwire_agents/schema.json

@@ -1925,9 +1925,17 @@
             {
               "type": "string",
               "const": "gpt-4.1-nano"
+            },
+            {
+              "type": "string",
+              "const": "nova-micro"
+            },
+            {
+              "type": "string",
+              "const": "nova-lite"
             }
           ],
-          "description": "The model to use for the AI. Allowed values are `gpt-4o-mini`, `gpt-4.1-mini`, and `gpt-4.1-nano`."
+          "description": "The model to use for the AI. Allowed values are `gpt-4o-mini`, `gpt-4.1-mini`, `gpt-4.1-nano`, `nova-micro`, and `nova-lite`."
         },
         "ai_volume": {
           "anyOf": [
@@ -4005,6 +4013,18 @@
           "maximum": 2,
           "description": "Aversion to repeating lines. Float value between -2.0 and 2.0. Positive values decrease the model's likelihood to repeat the same line verbatim."
         },
+        "model": {
+          "anyOf": [
+            {
+              "type": "string",
+              "enum": ["gpt-4o-mini", "gpt-4.1-mini", "gpt-4.1-nano", "nova-micro", "nova-lite"]
+            },
+            {
+              "$ref": "#/$defs/SWMLVar"
+            }
+          ],
+          "description": "The model to use for the post-prompt. Allowed values are `gpt-4o-mini`, `gpt-4.1-mini`, `gpt-4.1-nano`, `nova-micro`, and `nova-lite`."
+        },
         "text": {
           "type": "string",
           "description": "The instructions to send to the agent."
@@ -4084,6 +4104,18 @@
           "maximum": 2,
           "description": "Aversion to repeating lines. Float value between -2.0 and 2.0. Positive values decrease the model's likelihood to repeat the same line verbatim."
         },
+        "model": {
+          "anyOf": [
+            {
+              "type": "string",
+              "enum": ["gpt-4o-mini", "gpt-4.1-mini", "gpt-4.1-nano", "nova-micro", "nova-lite"]
+            },
+            {
+              "$ref": "#/$defs/SWMLVar"
+            }
+          ],
+          "description": "The model to use for the post-prompt. Allowed values are `gpt-4o-mini`, `gpt-4.1-mini`, `gpt-4.1-nano`, `nova-micro`, and `nova-lite`."
+        },
         "pom": {
           "type": "array",
           "items": {
@@ -4166,6 +4198,18 @@
           "maximum": 2,
           "description": "Aversion to repeating lines. Float value between -2.0 and 2.0. Positive values decrease the model's likelihood to repeat the same line verbatim."
         },
+        "model": {
+          "anyOf": [
+            {
+              "type": "string",
+              "enum": ["gpt-4o-mini", "gpt-4.1-mini", "gpt-4.1-nano", "nova-micro", "nova-lite"]
+            },
+            {
+              "$ref": "#/$defs/SWMLVar"
+            }
+          ],
+          "description": "The model to use for the prompt. Allowed values are `gpt-4o-mini`, `gpt-4.1-mini`, `gpt-4.1-nano`, `nova-micro`, and `nova-lite`."
+        },
         "text": {
           "type": "string",
           "description": "The instructions to send to the agent."
@@ -4249,6 +4293,18 @@
           "maximum": 2,
           "description": "Aversion to repeating lines. Float value between -2.0 and 2.0. Positive values decrease the model's likelihood to repeat the same line verbatim."
         },
+        "model": {
+          "anyOf": [
+            {
+              "type": "string",
+              "enum": ["gpt-4o-mini", "gpt-4.1-mini", "gpt-4.1-nano", "nova-micro", "nova-lite"]
+            },
+            {
+              "$ref": "#/$defs/SWMLVar"
+            }
+          ],
+          "description": "The model to use for the prompt. Allowed values are `gpt-4o-mini`, `gpt-4.1-mini`, `gpt-4.1-nano`, `nova-micro`, and `nova-lite`."
+        },
         "pom": {
           "type": "array",
           "items": {

signalwire_agents/search/__init__.py

@@ -68,6 +68,8 @@ if _SEARCH_AVAILABLE:
         from .index_builder import IndexBuilder
         from .search_engine import SearchEngine
         from .search_service import SearchService
+        from .models import MODEL_ALIASES, DEFAULT_MODEL, resolve_model_alias
+        from .migration import SearchIndexMigrator
         
         __all__ = [
             'preprocess_query',
@@ -75,7 +77,11 @@ if _SEARCH_AVAILABLE:
             'DocumentProcessor',
             'IndexBuilder',
             'SearchEngine',
-            'SearchService'
+            'SearchService',
+            'MODEL_ALIASES',
+            'DEFAULT_MODEL',
+            'resolve_model_alias',
+            'SearchIndexMigrator'
         ]
     except ImportError as e:
         # Some search components failed to import

signalwire_agents/search/document_processor.py

@@ -1075,7 +1075,7 @@ class DocumentProcessor:
                 json_metadata = json_chunk.get('metadata', {})
                 chunk_type = json_chunk.get('type', 'content')
                 
-                # Build chunk metadata
+                # Build chunk metadata (excluding tags which go at top level)
                 metadata = {
                     'chunk_method': 'json',
                     'chunk_index': idx,
@@ -1083,7 +1083,11 @@ class DocumentProcessor:
                     'original_chunk_id': json_chunk.get('chunk_id', f'chunk_{idx}')
                 }
                 
-                # Merge JSON metadata
+                # Extract tags before merging metadata
+                tags = json_metadata.get('tags', [])
+                
+                # Merge JSON metadata (this includes all fields including tags)
+                # We'll keep tags in metadata for backward compatibility but also set at top level
                 metadata.update(json_metadata)
                 
                 # Determine section name
@@ -1100,12 +1104,11 @@ class DocumentProcessor:
                     metadata=metadata
                 )
                 
-                # Add any additional fields from JSON
-                if 'tags' in json_chunk:
-                    chunk['tags'] = json_chunk['tags']
-                
-                # For TOC entries, we might want to add special tags
-                if chunk_type == 'toc' and 'tags' not in chunk:
+                # Set tags at the top level for proper tag filtering
+                if tags:
+                    chunk['tags'] = tags
+                elif chunk_type == 'toc':
+                    # For TOC entries, add special tags if none provided
                     chunk['tags'] = ['toc', 'navigation']
                 
                 chunks.append(chunk)

signalwire_agents/search/index_builder.py

@@ -85,9 +85,6 @@ class IndexBuilder:
         if self.backend not in ['sqlite', 'pgvector']:
             raise ValueError(f"Invalid backend '{self.backend}'. Must be 'sqlite' or 'pgvector'")
         
-        if self.backend == 'pgvector' and not self.connection_string:
-            raise ValueError("connection_string is required for pgvector backend")
-        
         # Validate NLP backend
         if self.index_nlp_backend not in ['nltk', 'spacy']:
             logger.warning(f"Invalid index_nlp_backend '{self.index_nlp_backend}', using 'nltk'")
@@ -105,6 +102,50 @@ class IndexBuilder:
             topic_threshold=self.topic_threshold
         )
     
+    def _extract_metadata_from_json_content(self, content: str) -> tuple[Dict[str, Any], str]:
+        """
+        Extract metadata from JSON content if present
+        
+        Returns:
+            (metadata_dict, metadata_text)
+        """
+        metadata_dict = {}
+        
+        # Try to extract metadata from JSON structure in content
+        if '"metadata":' in content:
+            try:
+                # Look for metadata object in content
+                import re
+                # Find all metadata objects
+                pattern = r'"metadata"\s*:\s*(\{[^{}]*(?:\{[^{}]*\}[^{}]*)*\})'
+                matches = re.finditer(pattern, content)
+                
+                for match in matches:
+                    try:
+                        json_metadata = json.loads(match.group(1))
+                        # Merge all found metadata
+                        if isinstance(json_metadata, dict):
+                            metadata_dict.update(json_metadata)
+                    except:
+                        pass
+            except Exception as e:
+                logger.debug(f"Error extracting JSON metadata: {e}")
+        
+        # Create searchable text from all metadata keys and values
+        metadata_text_parts = []
+        for key, value in metadata_dict.items():
+            # Add key
+            metadata_text_parts.append(str(key))
+            # Add value(s)
+            if isinstance(value, list):
+                metadata_text_parts.extend(str(v) for v in value)
+            else:
+                metadata_text_parts.append(str(value))
+        
+        metadata_text = ' '.join(metadata_text_parts).lower()
+        
+        return metadata_dict, metadata_text
+    
     def _load_model(self):
         """Load embedding model (lazy loading)"""
         if self.model is None:
@@ -147,6 +188,7 @@ class IndexBuilder:
         
         # Process documents
         chunks = []
+        print(f"Processing {len(files)} files...")
         for file_path in files:
             try:
                 # For individual files, use the file's parent as the base directory
@@ -154,8 +196,8 @@ class IndexBuilder:
                 base_dir = self._get_base_directory_for_file(file_path, sources)
                 file_chunks = self._process_file(file_path, base_dir, tags)
                 chunks.extend(file_chunks)
-                if self.verbose:
-                    print(f"Processed {file_path}: {len(file_chunks)} chunks")
+                if self.verbose or file_path.suffix == '.json':
+                    print(f"  {file_path}: {len(file_chunks)} chunks")
             except Exception as e:
                 logger.error(f"Error processing {file_path}: {e}")
                 if self.verbose:
@@ -171,7 +213,9 @@ class IndexBuilder:
         # Generate embeddings
         self._load_model()
         if self.verbose:
-            print("Generating embeddings...")
+            print(f"Generating embeddings for {len(chunks)} chunks...")
+        else:
+            print(f"Generating embeddings for {len(chunks)} chunks...")
         
         for i, chunk in enumerate(chunks):
             try:
@@ -183,15 +227,33 @@ class IndexBuilder:
                 )
                 
                 chunk['processed_content'] = processed['enhanced_text']
-                chunk['keywords'] = processed.get('keywords', [])
+                
+                # Include tags in keywords for better search matching
+                keywords = processed.get('keywords', [])
+                chunk_tags = chunk.get('tags', [])
+                if chunk_tags:
+                    # Add tags to keywords list for FTS matching
+                    keywords.extend(chunk_tags)
+                    # Remove duplicates while preserving order
+                    keywords = list(dict.fromkeys(keywords))
+                
+                chunk['keywords'] = keywords
+                
+                # For embedding, include tags in the text for better semantic matching
+                embedding_text = processed['enhanced_text']
+                if chunk_tags:
+                    # Append tags to the text for embedding generation
+                    embedding_text += " " + " ".join(chunk_tags)
                 
                 # Generate embedding (suppress progress bar)
-                embedding = self.model.encode(processed['enhanced_text'], show_progress_bar=False)
+                embedding = self.model.encode(embedding_text, show_progress_bar=False)
                 chunk['embedding'] = embedding.tobytes()
                 
-                if self.verbose and (i + 1) % 50 == 0:
+                # Show progress more frequently
+                show_every = 50 if len(chunks) > 500 else max(10, len(chunks) // 10)
+                if (i + 1) % show_every == 0 or (i + 1) == len(chunks):
                     progress_pct = ((i + 1) / len(chunks)) * 100
-                    print(f"Generated embeddings: {i + 1}/{len(chunks)} chunks ({progress_pct:.1f}%)")
+                    print(f"  Progress: {i + 1}/{len(chunks)} chunks ({progress_pct:.1f}%)")
                     
             except Exception as e:
                 logger.error(f"Error processing chunk {i}: {e}")
@@ -485,6 +547,7 @@ class IndexBuilder:
                     end_line INTEGER,
                     tags TEXT,
                     metadata TEXT,
+                    metadata_text TEXT,  -- Searchable text representation of all metadata
                     chunk_hash TEXT UNIQUE,
                     created_at TEXT DEFAULT CURRENT_TIMESTAMP
                 )
@@ -494,6 +557,7 @@ class IndexBuilder:
                 CREATE VIRTUAL TABLE chunks_fts USING fts5(
                     processed_content,
                     keywords,
+                    metadata_text,
                     content='chunks',
                     content_rowid='id'
                 )
@@ -555,13 +619,47 @@ class IndexBuilder:
                 # Prepare data
                 keywords_json = json.dumps(chunk.get('keywords', []))
                 tags_json = json.dumps(chunk.get('tags', []))
-                metadata_json = json.dumps(chunk.get('metadata', {}))
+                
+                # Extract metadata from JSON content and merge with chunk metadata
+                json_metadata, json_metadata_text = self._extract_metadata_from_json_content(chunk['content'])
+                chunk_metadata = chunk.get('metadata', {})
+                
+                # Merge metadata: chunk metadata takes precedence
+                merged_metadata = {**json_metadata, **chunk_metadata}
+                metadata_json = json.dumps(merged_metadata)
+                
+                # Create comprehensive metadata_text including tags
+                metadata_text_parts = []
+                
+                # Add metadata text from JSON content
+                if json_metadata_text:
+                    metadata_text_parts.append(json_metadata_text)
+                
+                # Add tags
+                tags = chunk.get('tags', [])
+                if tags:
+                    metadata_text_parts.extend(str(tag).lower() for tag in tags)
+                
+                # Add section if present
+                if chunk.get('section'):
+                    metadata_text_parts.append(chunk['section'].lower())
+                
+                # Add any additional metadata values
+                for key, value in chunk_metadata.items():
+                    if key not in json_metadata:  # Avoid duplicates
+                        metadata_text_parts.append(str(key).lower())
+                        if isinstance(value, list):
+                            metadata_text_parts.extend(str(v).lower() for v in value)
+                        else:
+                            metadata_text_parts.append(str(value).lower())
+                
+                metadata_text = ' '.join(metadata_text_parts)
                 
                 cursor.execute('''
                     INSERT OR IGNORE INTO chunks (
                         content, processed_content, keywords, language, embedding,
-                        filename, section, start_line, end_line, tags, metadata, chunk_hash
-                    ) VALUES (?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?)
+                        filename, section, start_line, end_line, tags, metadata, metadata_text, chunk_hash
+                    ) VALUES (?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?)
                 ''', (
                     chunk['content'],
                     chunk.get('processed_content', chunk['content']),
@@ -574,6 +672,7 @@ class IndexBuilder:
                     chunk.get('end_line'),
                     tags_json,
                     metadata_json,
+                    metadata_text,
                     chunk_hash
                 ))