Merge pull request #43 from kyaukyuai/feat/perplexity

✨ (search): integrate Perplexity API for enhanced search functionality

.env.example

@@ -9,6 +9,7 @@ ENVIRONMENT=development # development or production
 # AI Service Configuration
 ANTHROPIC_API_KEY=your-anthropic-api-key # Required for Claude AI integration
 TAVILY_API_KEY=your-tavily-api-key # Required for search functionality
+PERPLEXITY_API_KEY=your-perplexity-api-key # Required for search functionality
 
 # LangGraph Configuration
 LANGGRAPH_URL=http://localhost:2024 # LangGraph service endpoint

slack_ai_agent/agents/configuration.py

@@ -53,8 +53,8 @@ class Configuration:
     writer_provider: WriterProvider = (
         WriterProvider.ANTHROPIC
     )  # Defaults to Anthropic as provider
-    writer_model: str = "claude-3-5-sonnet-latest"  # Defaults to Anthropic as provider
-    search_api: SearchAPI = SearchAPI.TAVILY  # Default to TAVILY
+    writer_model: str = "claude-3-7-sonnet-latest"  # Defaults to Anthropic as provider
+    search_api: SearchAPI = SearchAPI.PERPLEXITY  # Default to TAVILY
 
     @classmethod
     def from_runnable_config(

slack_ai_agent/agents/deep_research_agent.py

@@ -36,6 +36,7 @@
 from slack_ai_agent.agents.prompts.deep_research_section_writer_instructions import (
     section_writer_instructions,
 )
+from slack_ai_agent.agents.tools.perplexity_search import perplexity_search
 from slack_ai_agent.agents.tools.tavily_search import deduplicate_and_format_sources
 from slack_ai_agent.agents.tools.tavily_search import tavily_search_async
 
@@ -197,11 +198,11 @@ async def generate_report_plan(state: ReportState, config: RunnableConfig):
         source_str = deduplicate_and_format_sources(
             search_results, max_tokens_per_source=1000, include_raw_content=False
         )
-    # elif search_api == "perplexity":
-    #     search_results = perplexity_search(query_list)
-    #     source_str = deduplicate_and_format_sources(
-    #         search_results, max_tokens_per_source=1000, include_raw_content=False
-    #     )
+    elif search_api == "perplexity":
+        search_results = perplexity_search(query_list)
+        source_str = deduplicate_and_format_sources(
+            search_results, max_tokens_per_source=1000, include_raw_content=False
+        )
     else:
         raise ValueError(f"Unsupported search API: {configurable.search_api}")
 

slack_ai_agent/agents/sync_deep_research_agent.py

@@ -36,6 +36,7 @@
 from slack_ai_agent.agents.prompts.deep_research_section_writer_instructions import (
     section_writer_instructions,
 )
+from slack_ai_agent.agents.tools.perplexity_search import perplexity_search
 from slack_ai_agent.agents.tools.tavily_search import deduplicate_and_format_sources
 from slack_ai_agent.agents.tools.tavily_search import (
     tavily_search,  # 非同期版ではなく同期版を使用
@@ -195,15 +196,15 @@ def generate_report_plan(state: ReportState, config: RunnableConfig):
 
     # Search the web - 同期バージョンを使用
     if search_api == "tavily":
-        search_results = tavily_search(query_list)  # 非同期から同期に変更
+        search_results = tavily_search(query=query_list)
+        source_str = deduplicate_and_format_sources(
+            search_results, max_tokens_per_source=1000, include_raw_content=False
+        )
+    elif search_api == "perplexity":
+        search_results = perplexity_search(search_queries=query_list)
         source_str = deduplicate_and_format_sources(
             search_results, max_tokens_per_source=1000, include_raw_content=False
         )
-    # elif search_api == "perplexity":
-    #     search_results = perplexity_search(query_list)
-    #     source_str = deduplicate_and_format_sources(
-    #         search_results, max_tokens_per_source=1000, include_raw_content=False
-    #     )
     else:
         raise ValueError(f"Unsupported search API: {configurable.search_api}")
 
@@ -375,15 +376,15 @@ def search_web(state: SectionState, config: RunnableConfig):
 
     # Search the web - 同期バージョン
     if search_api == "tavily":
-        search_results = tavily_search(query_list)  # 非同期から同期に変更
+        search_results = tavily_search(query=query_list)  # 非同期から同期に変更
         source_str = deduplicate_and_format_sources(
             search_results, max_tokens_per_source=5000, include_raw_content=True
         )
-    # elif search_api == "perplexity":
-    #     search_results = perplexity_search(query_list)
-    #     source_str = deduplicate_and_format_sources(
-    #         search_results, max_tokens_per_source=5000, include_raw_content=False
-    #     )
+    elif search_api == "perplexity":
+        search_results = perplexity_search(search_queries=query_list)
+        source_str = deduplicate_and_format_sources(
+            search_results, max_tokens_per_source=5000, include_raw_content=False
+        )
     else:
         raise ValueError(f"Unsupported search API: {configurable.search_api}")
 
@@ -431,7 +432,11 @@ def write_section(
     )
 
     # Write content to the section object
-    section.content = section_content.content
+    # Ensure we're getting a string from section_content
+    if hasattr(section_content, "content"):
+        section.content = str(section_content.content)
+    else:
+        section.content = str(section_content)
 
     # Grade prompt
     section_grader_instructions_formatted = section_grader_instructions.format(
@@ -449,16 +454,17 @@ def write_section(
         ]
     )
 
-    if (
-        feedback.grade == "pass"
-        or state["search_iterations"] >= configurable.max_search_depth
-    ):
+    # Safely access feedback attributes
+    grade = getattr(feedback, "grade", None)
+    follow_up_queries = getattr(feedback, "follow_up_queries", [])
+
+    if grade == "pass" or state["search_iterations"] >= configurable.max_search_depth:
         # Publish the section to completed sections
         return Command(update={"completed_sections": [section]}, goto=END)
     else:
         # Update the existing section with new content and update search queries
         return Command(
-            update={"search_queries": feedback.follow_up_queries, "section": section},
+            update={"search_queries": follow_up_queries, "section": section},
             goto="search_web",
         )
 
@@ -498,7 +504,11 @@ def write_final_sections(state: SectionState, config: RunnableConfig):
     )
 
     # Write content to section
-    section.content = section_content.content
+    # Ensure we're getting a string from section_content
+    if hasattr(section_content, "content"):
+        section.content = str(section_content.content)
+    else:
+        section.content = str(section_content)
 
     # Write the updated section to completed sections
     return {"completed_sections": [section]}

slack_ai_agent/agents/tools/perplexity_search.py

@@ -0,0 +1,100 @@
+import os
+
+import requests  # type: ignore
+from langsmith import traceable
+
+
+@traceable
+def perplexity_search(search_queries):
+    """Search the web using the Perplexity API.
+
+    Args:
+        search_queries (List[str]): List of search query strings to process
+
+    Returns:
+        List[dict]: List of search responses from Perplexity API, one per query. Each response has format:
+            {
+                'query': str,                    # The original search query
+                'follow_up_questions': None,
+                'answer': None,
+                'images': list,
+                'results': [                     # List of search results
+                    {
+                        'title': str,            # Title of the search result
+                        'url': str,              # URL of the result
+                        'content': str,          # Summary/snippet of content
+                        'score': float,          # Relevance score
+                        'raw_content': str|None  # Full content or None for secondary citations
+                    },
+                    ...
+                ]
+            }
+    """
+
+    headers = {
+        "accept": "application/json",
+        "content-type": "application/json",
+        "Authorization": f"Bearer {os.getenv('PERPLEXITY_API_KEY')}",
+    }
+
+    search_docs = []
+    for query in search_queries:
+        payload = {
+            "model": "sonar-pro",
+            "messages": [
+                {
+                    "role": "system",
+                    "content": "Search the web and provide factual information with sources.",
+                },
+                {"role": "user", "content": query},
+            ],
+        }
+
+        response = requests.post(
+            "https://api.perplexity.ai/chat/completions", headers=headers, json=payload
+        )
+        response.raise_for_status()  # Raise exception for bad status codes
+
+        # Parse the response
+        data = response.json()
+        content = data["choices"][0]["message"]["content"]
+        citations = data.get("citations", ["https://perplexity.ai"])
+
+        # Create results list for this query
+        results = []
+
+        # First citation gets the full content
+        results.append(
+            {
+                "title": "Perplexity Search, Source 1",
+                "url": citations[0],
+                "content": content,
+                "raw_content": content,
+                "score": 1.0,  # Adding score to match Tavily format
+            }
+        )
+
+        # Add additional citations without duplicating content
+        for i, citation in enumerate(citations[1:], start=2):
+            results.append(
+                {
+                    "title": f"Perplexity Search, Source {i}",
+                    "url": citation,
+                    "content": "See primary source for full content",
+                    "raw_content": None,
+                    "score": 0.5,  # Lower score for secondary sources
+                }
+            )
+
+        # Format response to match Tavily structure
+        search_docs.append(
+            {
+                "query": query,
+                "follow_up_questions": None,
+                "answer": None,
+                "images": [],
+                "results": results,
+            }
+        )
+
+    return search_docs