Merge pull request #817 from macrocosm-os/release/v3.0.6

- v3.0.6 release.
- Bump bittensor to v9.9.0 to address commit reveal error during weight set.
- Fix deep researcher final answer sometimes formatted as a JSON.

Author: dbobrenko

apex/services/deep_research/deep_research_langchain.py

@@ -171,7 +171,7 @@ async def invoke(
         collected_sources: list[dict[str, str]] = []
         seen_urls: set[str] = set()
 
-        agent_chain = self._build_agent_chain()
+        agent_chain = await self._build_agent_chain()
 
         while step_index < max_iterations:
             logger.debug(f"Starting deep researcher {step_index + 1}/{max_iterations} step")
@@ -202,17 +202,18 @@ async def invoke(
 
             # Final answer branch
             if "final_answer" in parsed:
-                logger.debug("Early-stopping deep research due to the final answer")
                 final_answer = str(parsed.get("final_answer", ""))
                 reasoning_traces.append(
                     {
                         "step": f"iteration-{step_index}",
                         "model": getattr(self.research_model, "model_name", "unknown"),
                         "thought": thought,
-                        "final_answer": final_answer,
+                        "observation": final_answer,
                     }
                 )
-                return final_answer, self.tool_history, reasoning_traces
+                logger.debug("Early-stopping deep research due to the final answer")
+                # return final_answer, self.tool_history, reasoning_traces
+                break
 
             # Action branch (only websearch supported)
             action = parsed.get("action") or {}
@@ -255,7 +256,7 @@ async def invoke(
                         "model": getattr(self.research_model, "model_name", "unknown"),
                         "thought": thought,
                         "action": {"tool": "websearch", "query": query, "max_results": max_results},
-                        "observation": observation_text[:1000],
+                        "observation": observation_text[:1200],
                     }
                 )
                 continue
@@ -289,7 +290,7 @@ async def invoke(
                         "model": getattr(self.research_model, "model_name", "unknown"),
                         "thought": thought,
                         "action": {"tool": "python_repl", "code": code[:1000]},
-                        "observation": observation_text[:1000],
+                        "observation": observation_text[:1200],
                     }
                 )
                 continue
@@ -305,19 +306,9 @@ async def invoke(
             )
             notes.append("Agent returned an unsupported action. Use the websearch tool or provide final_answer.")
 
-        # Fallback: if loop ends without final answer, ask final model to synthesize from notes
+        # If loop ends without final answer, ask final model to synthesize from notes.
         logger.debug("Generating final answer")
-        final_prompt = PromptTemplate(
-            input_variables=["question", "notes", "sources"],
-            template=(
-                self._FINAL_ANSWER_INST + "Do NOT use JSON, or any other structured data format.\n"
-                "Question:\n{question}\n\n"
-                "Notes:\n{notes}\n\n"
-                "Sources:\n{sources}\n\n"
-                "Research Report:"
-            ),
-        )
-        final_chain = final_prompt | self.final_model | StrOutputParser()
+        final_chain = await self._build_final_chain()
 
         final_report: str = await self._try_invoke(
             final_chain,
@@ -336,6 +327,19 @@ async def invoke(
         )
         return final_report, self.tool_history, reasoning_traces
 
+    async def _build_final_chain(self) -> RunnableSerializable[dict[str, Any], str]:
+        final_prompt = PromptTemplate(
+            input_variables=["question", "notes", "sources"],
+            template=(
+                self._FINAL_ANSWER_INST + "Do NOT use JSON, or any other structured data format. Provide \n"
+                "Question:\n{question}\n\n"
+                "Notes:\n{notes}\n\n"
+                "Sources:\n{sources}\n\n"
+                "Research Report:"
+            ),
+        )
+        return final_prompt | self.final_model | StrOutputParser()
+
     def _render_sources(self, collected_sources: list[dict[str, str]], max_items: int = 12) -> str:
         if not collected_sources:
             return "(none)"
@@ -352,47 +356,49 @@ def _render_notes(self, notes: list[str], max_items: int = 8) -> str:
         clipped = notes[-max_items:]
         return "\n".join(f"- {item}" for item in clipped)
 
-    def _build_agent_chain(self) -> RunnableSerializable[dict[str, Any], str]:
+    async def _build_agent_chain(self) -> RunnableSerializable[dict[str, Any], str]:
         prompt = PromptTemplate(
             input_variables=["question", "notes", "sources"],
             template=(
                 "You are DeepResearcher, a meticulous, tool-using research agent.\n"
                 "You can use exactly these tools: websearch, python_repl.\n\n"
                 "Tool: websearch\n"
-                "- description: Search the web for relevant information.\n"
-                "- args: keys: 'query' (string), 'max_results' (integer <= 10)\n\n"
+                "  - description: Search the web for relevant information.\n"
+                "  - args: keys: 'query' (string), 'max_results' (integer <= 10)\n\n"
                 "Tool: python_repl\n"
-                "- description: A Python shell for executing Python commands.\n"
-                "- note: Print values to see output, e.g., `print(...)`.\n"
-                "- args: keys: 'code' (string: valid python command).\n\n"
+                "  - description: A Python shell for executing Python commands.\n"
+                "  - note: Print values to see output, e.g., `print(...)`.\n"
+                "  - args: keys: 'code' (string: valid python command).\n\n"
                 "Follow an iterative think-act-observe loop. "
                 "Prefer rich internal reasoning over issuing many tool calls.\n"
                 "Spend time thinking: produce substantial, explicit reasoning in each 'thought'.\n"
                 "Avoid giving a final answer too early. Aim for at least 6 detailed thoughts before finalizing,\n"
                 "unless the question is truly trivial. "
                 "If no tool use is needed in a step, still provide a reflective 'thought'\n"
                 "that evaluates evidence, identifies gaps, and plans the next step.\n\n"
-                "Always respond in strict JSON. Use one of the two schemas:\n\n"
-                "1) Action step (JSON keys shown with dot-paths):\n"
-                "- thought: string\n"
-                "- action.tool: 'websearch' | 'python_repl'\n"
-                "- action.input: for websearch -> {{query: string, max_results: integer}}\n"
-                "- action.input: for python_repl -> {{code: string}}\n\n"
-                "2) Final answer step:\n"
-                "- thought: string\n"
-                "- final_answer: string (use plain text for final answer, not a JSON)\n\n"
+                "Always respond in strict JSON for deep research steps (do not use JSON for final answer). "
+                "Use one of the two schemas:\n\n"
+                "1. Action step (JSON keys shown with dot-paths):\n"
+                "  - thought: string\n"
+                "  - action.tool: 'websearch' | 'python_repl'\n"
+                "  - action.input: for websearch -> {{query: string, max_results: integer}}\n"
+                "  - action.input: for python_repl -> {{code: string}}\n\n"
+                "2. Final answer step:\n"
+                "  - thought: string\n"
+                "  - final_answer: string\n\n"
                 "In every step, make 'thought' a detailed paragraph (120-200 words) that:\n"
-                "- Summarizes what is known and unknown so far\n"
-                "- Justifies the chosen next action or decision not to act\n"
-                "- Evaluates evidence quality and cites source numbers when applicable\n"
-                "- Identifies risks, uncertainties, and alternative hypotheses\n\n"
+                "  - Summarizes what is known and unknown so far\n"
+                "  - Justifies the chosen next action or decision not to act\n"
+                "  - Evaluates evidence quality and cites source numbers when applicable\n"
+                "  - Identifies risks, uncertainties, and alternative hypotheses\n\n"
+                "Respond with JSON only during deep research steps, "
+                "final answer must be always in a plain text formatted as a research report, with sections:\n"
                 "Executive Summary, Key Findings, Evidence, Limitations, Conclusion.\n"
                 "Use inline numeric citations like [1], [2] that refer to Sources.\n"
                 "Include a final section titled 'Sources' listing the numbered citations.\n\n"
                 "Question:\n{question}\n\n"
                 "Notes and observations so far:\n{notes}\n\n"
                 "Sources (use these for citations):\n{sources}\n\n"
-                "Respond with JSON always, except for final_anwer (use plain text)."
             ),
         )
         return prompt | self.research_model | StrOutputParser()

pyproject.toml

@@ -1,6 +1,6 @@
 [project]
 name = "apex"
-version = "3.0.5"
+version = "3.0.6"
 description = "Bittensor Subnet 1: Apex"
 readme = "README.md"
 requires-python = "~=3.11"
@@ -26,7 +26,7 @@ dependencies = [
     "loguru>=0.7.3",
     "tavily-python>=0.7.10",
     "pip>=25.1.1",
-    "bittensor>=9.8.3",
+    "bittensor==9.9.0",
     "rouge>=1.0.1",
     "substrate-interface>=1.7.11",
     "types-netaddr>=1.3.0.20240530",

tests/services/deep_research/test_deep_research_langchain.py

@@ -1,3 +1,4 @@
+import json
 from unittest.mock import AsyncMock, MagicMock, patch
 
 import pytest
@@ -76,17 +77,26 @@ async def test_invoke_with_documents_in_body(deep_research_langchain, mock_webse
     body = {"documents": [{"page_content": "doc1"}, {"page_content": "doc2"}]}
 
     with (
-        patch("apex.services.deep_research.deep_research_langchain.PromptTemplate") as mock_prompt_template,
-        patch("apex.services.deep_research.deep_research_langchain.StrOutputParser"),
+        patch(
+            "apex.services.deep_research.deep_research_langchain.DeepResearchLangchain._build_agent_chain"
+        ) as mock_build_agent_chain,
+        patch(
+            "apex.services.deep_research.deep_research_langchain.DeepResearchLangchain._build_final_chain"
+        ) as mock_build_final_chain,
     ):
         agent_chain = AsyncMock()
-        agent_chain.ainvoke.return_value = '{"thought": "enough info", "final_answer": "final_report"}'
-        mock_prompt_template.return_value.__or__.return_value.__or__.return_value = agent_chain
+        return_value = json.dumps({"thought": "enough info", "final_answer": "final_report"})
+        agent_chain.ainvoke.return_value = return_value
+        mock_build_agent_chain.return_value = agent_chain
+
+        final_chain_mock = AsyncMock()
+        final_chain_mock.ainvoke.return_value = return_value
+        mock_build_final_chain.return_value = final_chain_mock
 
         result = await deep_research_langchain.invoke(messages, body)
 
         mock_websearch.search.assert_not_called()
-        assert result[0] == "final_report"
+        assert result[0] == return_value
 
 
 @pytest.mark.asyncio
@@ -96,8 +106,12 @@ async def test_invoke_with_websearch(deep_research_langchain, mock_websearch):
     mock_websearch.search.return_value = [MagicMock(content="web_doc", url="http://a.com", title="A")]
 
     with (
-        patch("apex.services.deep_research.deep_research_langchain.PromptTemplate") as mock_prompt_template,
-        patch("apex.services.deep_research.deep_research_langchain.StrOutputParser"),
+        patch(
+            "apex.services.deep_research.deep_research_langchain.DeepResearchLangchain._build_agent_chain"
+        ) as mock_build_agent_chain,
+        patch(
+            "apex.services.deep_research.deep_research_langchain.DeepResearchLangchain._build_final_chain"
+        ) as mock_build_final_chain,
     ):
         agent_chain = AsyncMock()
         agent_chain.ainvoke.side_effect = [
@@ -107,7 +121,11 @@ async def test_invoke_with_websearch(deep_research_langchain, mock_websearch):
             ),
             '{"thought": "done", "final_answer": "final_answer"}',
         ]
-        mock_prompt_template.return_value.__or__.return_value.__or__.return_value = agent_chain
+        mock_build_agent_chain.return_value = agent_chain
+
+        final_chain_mock = AsyncMock()
+        final_chain_mock.ainvoke.return_value = "final_answer"
+        mock_build_final_chain.return_value = final_chain_mock
 
         result = await deep_research_langchain.invoke(messages)
 
@@ -121,17 +139,26 @@ async def test_invoke_no_websearch_needed_final_answer(deep_research_langchain,
     messages = [{"role": "user", "content": "test question"}]
 
     with (
-        patch("apex.services.deep_research.deep_research_langchain.PromptTemplate") as mock_prompt_template,
-        patch("apex.services.deep_research.deep_research_langchain.StrOutputParser"),
+        patch(
+            "apex.services.deep_research.deep_research_langchain.DeepResearchLangchain._build_agent_chain"
+        ) as mock_build_agent_chain,
+        patch(
+            "apex.services.deep_research.deep_research_langchain.DeepResearchLangchain._build_final_chain"
+        ) as mock_build_final_chain,
     ):
         agent_chain = AsyncMock()
-        agent_chain.ainvoke.return_value = '{"thought": "clear", "final_answer": "final_report"}'
-        mock_prompt_template.return_value.__or__.return_value.__or__.return_value = agent_chain
+        return_value = json.dumps({"thought": "enough info", "final_answer": "final_report"})
+        agent_chain.ainvoke.return_value = return_value
+        mock_build_agent_chain.return_value = agent_chain
+
+        final_chain_mock = AsyncMock()
+        final_chain_mock.ainvoke.return_value = return_value
+        mock_build_final_chain.return_value = final_chain_mock
 
         result = await deep_research_langchain.invoke(messages)
 
         mock_websearch.search.assert_not_called()
-        assert result[0] == "final_report"
+        assert result[0] == return_value
 
 
 @pytest.mark.asyncio
@@ -149,8 +176,12 @@ async def test_full_invoke_flow_with_multiple_actions(deep_research_langchain, m
     ]
 
     with (
-        patch("apex.services.deep_research.deep_research_langchain.PromptTemplate") as mock_prompt_template,
-        patch("apex.services.deep_research.deep_research_langchain.StrOutputParser"),
+        patch(
+            "apex.services.deep_research.deep_research_langchain.DeepResearchLangchain._build_agent_chain"
+        ) as mock_build_agent_chain,
+        patch(
+            "apex.services.deep_research.deep_research_langchain.DeepResearchLangchain._build_final_chain"
+        ) as mock_build_final_chain,
     ):
         agent_chain = AsyncMock()
         agent_chain.ainvoke.side_effect = [
@@ -164,7 +195,11 @@ async def test_full_invoke_flow_with_multiple_actions(deep_research_langchain, m
             ),
             '{"thought": "complete", "final_answer": "final_report"}',
         ]
-        mock_prompt_template.return_value.__or__.return_value.__or__.return_value = agent_chain
+        mock_build_agent_chain.return_value = agent_chain
+
+        final_chain_mock = AsyncMock()
+        final_chain_mock.ainvoke.return_value = "final_report"
+        mock_build_final_chain.return_value = final_chain_mock
 
         result = await deep_research_langchain.invoke(messages)
 
@@ -186,15 +221,23 @@ async def test_full_invoke_flow_with_multiple_actions(deep_research_langchain, m
 async def test_invoke_with_python_repl(deep_research_langchain):
     """Agent chooses python_repl then produces final answer."""
     with (
-        patch("apex.services.deep_research.deep_research_langchain.PromptTemplate") as mock_prompt_template,
-        patch("apex.services.deep_research.deep_research_langchain.StrOutputParser"),
+        patch(
+            "apex.services.deep_research.deep_research_langchain.DeepResearchLangchain._build_agent_chain"
+        ) as mock_build_agent_chain,
+        patch(
+            "apex.services.deep_research.deep_research_langchain.DeepResearchLangchain._build_final_chain"
+        ) as mock_build_final_chain,
     ):
         agent_chain = AsyncMock()
         agent_chain.ainvoke.side_effect = [
             ('{"thought": "compute needed", "action": {"tool": "python_repl", "input": {"code": "print(1+1)"}}}'),
             '{"thought": "done", "final_answer": "final_answer"}',
         ]
-        mock_prompt_template.return_value.__or__.return_value.__or__.return_value = agent_chain
+        mock_build_agent_chain.return_value = agent_chain
+
+        final_chain_mock = AsyncMock()
+        final_chain_mock.ainvoke.return_value = "final_answer"
+        mock_build_final_chain.return_value = final_chain_mock
 
         result = await deep_research_langchain.invoke([{"role": "user", "content": "q"}])