Feature/improved detailed report

backend/report_type/detailed_report/detailed_report.py

@@ -3,7 +3,7 @@
 from fastapi import WebSocket
 
 from gpt_researcher.master.actions import (
-    add_source_urls,
+    add_references,
     extract_headers,
     extract_sections,
     table_of_contents,
@@ -17,10 +17,10 @@ def __init__(
         query: str,
         report_type: str,
         report_source: str,
-        source_urls: List[str],
-        config_path: str,
-        tone: Tone,
-        websocket: WebSocket,
+        source_urls: List[str] = [],
+        config_path: str = None,
+        tone: Tone = Tone.Formal,
+        websocket: WebSocket = None,
         subtopics: List[Dict] = [],
         headers: Optional[Dict] = None
     ):
@@ -121,5 +121,6 @@ async def _get_subtopic_report(self, subtopic: Dict) -> Dict[str, str]:
 
     async def _construct_detailed_report(self, introduction: str, report_body: str) -> str:
         toc = table_of_contents(report_body)
-        report_with_references = add_source_urls(report_body, self.main_task_assistant.visited_urls)
-        return f"{introduction}\n\n{toc}\n\n{report_with_references}"
+        conclusion = await self.main_task_assistant.write_report_conclusion(report_body)
+        conclusion_with_references = add_references(conclusion, self.main_task_assistant.visited_urls)
+        return f"{introduction}\n\n{toc}\n\n{report_body}\n\n{conclusion_with_references}"

cli.py

@@ -17,6 +17,7 @@
 
 from gpt_researcher import GPTResearcher
 from gpt_researcher.utils.enum import ReportType
+from backend.report_type import DetailedReport
 
 # =============================================================================
 # CLI
@@ -45,6 +46,7 @@
 
 report_type_descriptions = {
     ReportType.ResearchReport.value: "Summary - Short and fast (~2 min)",
+    ReportType.DetailedReport.value: "Detailed - In depth and longer (~5 min)",
     ReportType.ResourceReport.value: "",
     ReportType.OutlineReport.value: "",
     ReportType.CustomReport.value: "",
@@ -71,13 +73,22 @@ async def main(args):
     Conduct research on the given query, generate the report, and write
     it as a markdown file to the output directory.
     """
-    researcher = GPTResearcher(
-        query=args.query,
-        report_type=args.report_type)
-
-    await researcher.conduct_research()
-
-    report = await researcher.write_report()
+    if args.report_type == 'detailed_report':
+        detailed_report = DetailedReport(
+            query=args.query,
+            report_type="research_report",
+            report_source="web_search",
+        )
+
+        report = await detailed_report.run()
+    else:
+        researcher = GPTResearcher(
+            query=args.query,
+            report_type=args.report_type)
+
+        await researcher.conduct_research()
+
+        report = await researcher.write_report()
 
     # Write the report to a file
     artifact_filepath = f"outputs/{uuid4()}.md"

docs/docs/examples/detailed_report.md

@@ -49,7 +49,7 @@ Here's how you can use the `DetailedReport` class in your project:
 import asyncio
 from fastapi import WebSocket
 from gpt_researcher.utils.enum import Tone
-from backend.report_types.detailed_report import DetailedReport
+from backend.report_type import DetailedReport
 
 async def generate_report(websocket: WebSocket):
     detailed_report = DetailedReport(

gpt_researcher/master/actions.py

@@ -142,7 +142,7 @@ async def choose_agent(
                 {"role": "system", "content": f"{auto_agent_instructions()}"},
                 {"role": "user", "content": f"task: {query}"},
             ],
-            temperature=0,
+            temperature=0.15,
             llm_provider=cfg.llm_provider,
             llm_kwargs=cfg.llm_kwargs,
             cost_callback=cost_callback,
@@ -223,7 +223,7 @@ async def get_sub_queries(
                 ),
             },
         ],
-        temperature=0,
+        temperature=0.1,
         llm_provider=cfg.llm_provider,
         llm_kwargs=cfg.llm_kwargs,
         cost_callback=cost_callback,
@@ -318,6 +318,30 @@ def chunk_content(raw_content, chunk_size=10000):
 
     return concatenated_summaries
 
+async def write_conclusion(
+    report, agent_role_prompt, cfg, cost_callback: callable = None
+):
+    conclusion_prompt = generate_report_conclusion(report_content=report)
+    conclusion = ""
+    try:
+        conclusion = await create_chat_completion(
+            model=cfg.fast_llm_model,
+            messages=[
+                {"role": "system", "content": f"{agent_role_prompt}"},
+                {
+                    "role": "user",
+                    "content": f"{conclusion_prompt}",
+                },
+            ],
+            temperature=0.35,
+            llm_provider=cfg.llm_provider,
+            llm_kwargs=cfg.llm_kwargs,
+            cost_callback=cost_callback,
+        )
+    except Exception as e:
+        print(f"{Fore.RED}Error in generating report conclusion: {e}{Style.RESET_ALL}")
+    return conclusion
+
 
 async def summarize_url(
     query, raw_data, agent_role_prompt, cfg, cost_callback: callable = None
@@ -346,7 +370,7 @@ async def summarize_url(
                     "content": f"{generate_summary_prompt(query, raw_data)}",
                 },
             ],
-            temperature=0,
+            temperature=0.35,
             llm_provider=cfg.llm_provider,
             llm_kwargs=cfg.llm_kwargs,
             cost_callback=cost_callback,
@@ -375,7 +399,7 @@ async def generate_draft_section_titles(
                 {"role": "system", "content": f"{agent_role_prompt}"},
                 {"role": "user", "content": content},
             ],
-            temperature=0,
+            temperature=0.15,
             llm_provider=cfg.llm_provider,
             llm_kwargs=cfg.llm_kwargs,
             cost_callback=cost_callback,
@@ -433,7 +457,7 @@ async def generate_report(
                 {"role": "system", "content": f"{agent_role_prompt}"},
                 {"role": "user", "content": content},
             ],
-            temperature=0,
+            temperature=0.35,
             llm_provider=cfg.llm_provider,
             stream=True,
             websocket=websocket,
@@ -486,7 +510,7 @@ async def get_report_introduction(
                     "content": generate_report_introduction(query, context),
                 },
             ],
-            temperature=0,
+            temperature=0.25,
             llm_provider=config.llm_provider,
             stream=True,
             websocket=websocket,
@@ -600,7 +624,7 @@ def generate_table_of_contents(headers, indent_level=0):
         return markdown_text  # Return original markdown text if an exception occurs
 
 
-def add_source_urls(report_markdown: str, visited_urls: set):
+def add_references(report_markdown: str, visited_urls: set):
     """
     This function takes a Markdown report and a set of visited URLs as input parameters.
 

gpt_researcher/master/agent.py

@@ -482,6 +482,40 @@ def get_costs(self) -> float:
     def set_verbose(self, verbose: bool):
         self.verbose = verbose
 
+    async def write_report_conclusion(self, report_body: str) -> str:
+        """
+        Writes the conclusion of the report based on the research conducted.
+
+        Args:
+            report_body (str): The body of the report.
+
+        Returns:
+            str: The conclusion of the report.
+        """
+        if self.verbose:
+            await stream_output(
+                "logs",
+                "writing_conclusion",
+                f"🙇️ Concluding report for research task: {self.query}...",
+                self.websocket,
+            )
+
+        conclusion = await write_conclusion(
+            report=report_body,
+            agent_role_prompt=self.role,
+            cfg=self.cfg,
+        )
+
+        if self.verbose:
+            await stream_output(
+                "logs",
+                "report_conclusion",
+                f"✍️ Writing final conclusion: {conclusion}...",
+                self.websocket,
+            )
+
+        return conclusion
+
     def add_costs(self, cost: int) -> None:
         if not isinstance(cost, float) and not isinstance(cost, int):
             raise ValueError("Cost must be an integer or float")
@@ -492,6 +526,13 @@ def add_costs(self, cost: int) -> None:
     # DETAILED REPORT
 
     async def write_introduction(self):
+        if self.verbose:
+            await stream_output(
+                "logs",
+                "generating_conclusion",
+                f"🤔 Generating subtopics...",
+                self.websocket,
+            )
         # Construct Report Introduction from main topic research
         introduction = await get_report_introduction(
             self.query,

gpt_researcher/master/prompts.py

@@ -345,6 +345,26 @@ def generate_report_introduction(question: str, research_summary: str = "") -> s
 Assume that the current date is {datetime.now(timezone.utc).strftime('%B %d, %Y')} if required.
 """
 
+def generate_report_conclusion(report_content: str) -> str:
+    prompt = f"""
+    Based on the following research report, please write a concise conclusion that summarizes the main findings and their implications:
+
+    {report_content}
+
+    Your conclusion should:
+    1. Recap the main points of the research
+    2. Highlight the most important findings
+    3. Discuss any implications or next steps
+    4. Be approximately 2-3 paragraphs long
+
+    If there is no "## Conclusion" section title written at the end of the report, please add it to the top of your conclusion. 
+    You must include hyperlinks with markdown syntax ([url website](url)) related to the sentences wherever necessary.
+
+    Write the conclusion:
+    """
+
+    return prompt
+
 
 report_type_mapping = {
     ReportType.ResearchReport.value: generate_report_prompt,

setup.py

@@ -1,6 +1,6 @@
 from setuptools import find_packages, setup
 
-LATEST_VERSION = "0.9.4"
+LATEST_VERSION = "0.9.5"
 
 exclude_packages = [
     "selenium",