plutoa386 commited on
Commit
6efe67f
Β·
1 Parent(s): c8624cd

updated new gathering logics

Browse files
Files changed (1) hide show
  1. crew.py +49 -39
crew.py CHANGED
@@ -299,10 +299,9 @@ def explain_dementia_risk_lifestyle(docID: str) -> str:
299
  return f"SHAP explanation failed: {str(e)}"
300
 
301
  @tool("gather_news_patient")
302
- def gather_news_patient(count: int = 5) -> str:
303
  """
304
- Gathers recent dementia news articles for patients. Returns JSON-like string of raw results.
305
- Args: count (int): Number of articles to fetch (max 10).
306
  """
307
  from webscraper import search_dementia_news, scrape_article_content
308
  articles = search_dementia_news(audience="patient", max_results=min(count, 10))
@@ -314,18 +313,14 @@ def gather_news_patient(count: int = 5) -> str:
314
  clean_articles.append({
315
  "title": art["title"],
316
  "url": art["url"],
317
- "content": content[:3000]
318
  })
319
  except Exception:
320
  continue
321
- return json.dumps(clean_articles)
322
-
323
 
324
  @tool("gather_news_doctor")
325
- def gather_news_doctor(count: int = 5) -> str:
326
- """
327
- Gathers recent dementia news articles for doctors. Returns JSON-like string of raw results.
328
- """
329
  from webscraper import search_dementia_news, scrape_article_content
330
  articles = search_dementia_news(audience="doctor", max_results=min(count, 10))
331
  clean_articles = []
@@ -340,7 +335,7 @@ def gather_news_doctor(count: int = 5) -> str:
340
  })
341
  except Exception:
342
  continue
343
- return json.dumps(clean_articles)
344
 
345
  # Load 2 different llms for different purposes.
346
  llm_predict = LLM(model="huggingface/openai/gpt-oss-120b")
@@ -410,26 +405,38 @@ explanation_task_lifestyle = Task(
410
 
411
  get_news_task_patient = Task(
412
  description=(
413
- "1. Use gather_news_patient tool to fetch recent dementia news for patients.\n"
414
- "2. For each article, read its content and generate a concise 2–3 line summary highlighting key takeaways for patients/families.\n"
415
- "3. Ensure summaries are empathetic, jargon-free, and actionable."
 
 
 
 
416
  ),
417
  expected_output=(
418
  "A JSON object with keys: 'titles', 'snippets', 'urls'.\n"
419
- "- 'titles': list of article titles (strings).\n"
420
  "- 'snippets': list of 2–3 line plain-language summaries (strings), one per article.\n"
421
  "- 'urls': list of original article URLs (strings).\n"
422
  "All lists must be same length and aligned by index.\n"
423
- "Return ONLY valid JSON β€” no markdown, no extra text."
424
  ),
425
  agent=patient_news
426
  )
427
 
428
  get_news_task_doctor = Task(
429
  description=(
430
- "1. Use gather_news_doctor tool to fetch recent dementia news for clinicians.\n"
431
- "2. For each article, read its content and generate a concise 2–3 line summary highlighting clinical relevance (e.g., trial outcomes, guideline changes, diagnostic advances).\n"
432
- "3. Keep summaries professional and evidence-focused."
 
 
 
 
 
 
 
 
433
  ),
434
  expected_output=(
435
  "A JSON object with keys: 'titles', 'snippets', 'urls'.\n"
@@ -679,16 +686,10 @@ def run_crewai_audio(docID: str) -> dict:
679
 
680
  def run_crewai_news_patient(count: int = 5) -> dict:
681
  try:
682
- # Inject count into task description dynamically
683
- desc = get_news_task_patient.description.replace(
684
- "gather_news_patient tool",
685
- f"gather_news_patient tool with count={count}"
686
- )
687
- task = Task(description=desc, expected_output=get_news_task_patient.expected_output, agent=patient_news)
688
-
689
  crew = Crew(
690
  agents=[patient_news],
691
- tasks=[task],
692
  verbose=True,
693
  function_calling=False
694
  )
@@ -701,7 +702,7 @@ def run_crewai_news_patient(count: int = 5) -> dict:
701
  raise ValueError(f"Missing or invalid '{key}' in result")
702
  if not (len(result["titles"]) == len(result["snippets"]) == len(result["urls"])):
703
  raise ValueError("Mismatched list lengths")
704
-
705
  return {"status": "success", **result}
706
 
707
  except Exception as e:
@@ -709,29 +710,38 @@ def run_crewai_news_patient(count: int = 5) -> dict:
709
 
710
  def run_crewai_news_doctor(count: int = 5) -> dict:
711
  try:
712
- desc = get_news_task_doctor.description.replace(
713
- "gather_news_doctor tool",
714
- f"gather_news_doctor tool with count={count}"
715
- )
716
- task = Task(description=desc, expected_output=get_news_task_doctor.expected_output, agent=doctor_news)
717
 
 
 
 
 
 
 
 
 
718
  crew = Crew(
719
  agents=[doctor_news],
720
- tasks=[task],
721
  verbose=True,
722
  function_calling=False
723
  )
724
- raw = str(crew.kickoff()).strip()
725
 
726
- result = _parse_json_output(raw)
 
 
 
 
 
 
727
  for key in ["titles", "snippets", "urls"]:
728
  if key not in result or not isinstance(result[key], list):
729
- raise ValueError(f"Missing or invalid '{key}'")
730
  if not (len(result["titles"]) == len(result["snippets"]) == len(result["urls"])):
731
- raise ValueError("Mismatched list lengths")
732
 
733
  return {"status": "success", **result}
734
 
735
  except Exception as e:
736
  return {"status": "error", "message": str(e)}
737
- return {"type": "placeholder", "value": "jharden13 is tha GOAT..........doctor"}
 
299
  return f"SHAP explanation failed: {str(e)}"
300
 
301
  @tool("gather_news_patient")
302
+ def gather_news_patient(count: int = 5) -> dict: # ← Change return type to dict
303
  """
304
+ Gathers recent dementia news articles for patients. Returns structured dict.
 
305
  """
306
  from webscraper import search_dementia_news, scrape_article_content
307
  articles = search_dementia_news(audience="patient", max_results=min(count, 10))
 
313
  clean_articles.append({
314
  "title": art["title"],
315
  "url": art["url"],
316
+ "content": content[:3000] # Keep full content for LLM to summarize
317
  })
318
  except Exception:
319
  continue
320
+ return {"articles": clean_articles} # ← Return dict, not JSON string
 
321
 
322
  @tool("gather_news_doctor")
323
+ def gather_news_doctor(count: int = 5) -> dict: # ← Same here
 
 
 
324
  from webscraper import search_dementia_news, scrape_article_content
325
  articles = search_dementia_news(audience="doctor", max_results=min(count, 10))
326
  clean_articles = []
 
335
  })
336
  except Exception:
337
  continue
338
+ return {"articles": clean_articles}
339
 
340
  # Load 2 different llms for different purposes.
341
  llm_predict = LLM(model="huggingface/openai/gpt-oss-120b")
 
405
 
406
  get_news_task_patient = Task(
407
  description=(
408
+ "1. Use the 'gather_news_patient' tool to fetch recent dementia news articles.\n"
409
+ "2. You will receive a list of articles, each with 'title', 'url', and 'content'.\n"
410
+ "3. For EACH article, read its 'content' field and generate a concise 2–3 line summary that highlights key takeaways for patients/families.\n"
411
+ "4. Ensure summaries are empathetic, jargon-free, and actionable.\n"
412
+ "5. Return a JSON object with keys: 'titles', 'snippets', 'urls' β€” matching the order of the fetched articles.\n"
413
+ "6. DO NOT invent titles or URLs β€” use only what was returned by the tool.\n"
414
+ "7. Return ONLY valid JSON β€” no markdown, no extra text."
415
  ),
416
  expected_output=(
417
  "A JSON object with keys: 'titles', 'snippets', 'urls'.\n"
418
+ "- 'titles': list of original article titles (strings).\n"
419
  "- 'snippets': list of 2–3 line plain-language summaries (strings), one per article.\n"
420
  "- 'urls': list of original article URLs (strings).\n"
421
  "All lists must be same length and aligned by index.\n"
422
+ "Return ONLY valid JSON β€” no extra text."
423
  ),
424
  agent=patient_news
425
  )
426
 
427
  get_news_task_doctor = Task(
428
  description=(
429
+ "1. Use the 'gather_news_doctor' tool to fetch recent dementia-related articles for medical professionals.\n"
430
+ "2. You will receive a list of articles, each with 'title', 'url', and 'content' (scraped full text).\n"
431
+ "3. For EACH article, read its 'content' field and generate a concise 2–3 line summary.\n"
432
+ "4. Summaries must be clinical, evidence-focused, and highlight:\n"
433
+ " - Trial design/outcomes (if applicable)\n"
434
+ " - Diagnostic/therapeutic implications\n"
435
+ " - Guideline updates or expert consensus changes\n"
436
+ " - Biomarker or mechanistic insights\n"
437
+ "5. DO NOT invent titles, URLs, or content β€” use only what the tool returns.\n"
438
+ "6. Return a JSON object with keys: 'titles', 'snippets', 'urls', in matching order.\n"
439
+ "7. Return ONLY valid JSON β€” no markdown, no commentary."
440
  ),
441
  expected_output=(
442
  "A JSON object with keys: 'titles', 'snippets', 'urls'.\n"
 
686
 
687
  def run_crewai_news_patient(count: int = 5) -> dict:
688
  try:
689
+ # Just use the pre-defined task β€” it already knows to use gather_news_patient
 
 
 
 
 
 
690
  crew = Crew(
691
  agents=[patient_news],
692
+ tasks=[get_news_task_patient],
693
  verbose=True,
694
  function_calling=False
695
  )
 
702
  raise ValueError(f"Missing or invalid '{key}' in result")
703
  if not (len(result["titles"]) == len(result["snippets"]) == len(result["urls"])):
704
  raise ValueError("Mismatched list lengths")
705
+
706
  return {"status": "success", **result}
707
 
708
  except Exception as e:
 
710
 
711
  def run_crewai_news_doctor(count: int = 5) -> dict:
712
  try:
713
+ original_gather = doctor_news.tools[0]
 
 
 
 
714
 
715
+ # Wrap tool to inject count
716
+ def gather_with_count(query=None):
717
+ return gather_news_doctor(count=count)
718
+
719
+ # Replace tool temporarily
720
+ doctor_news.tools = [gather_with_count]
721
+
722
+ # Run crew with static task
723
  crew = Crew(
724
  agents=[doctor_news],
725
+ tasks=[get_news_task_doctor],
726
  verbose=True,
727
  function_calling=False
728
  )
729
+ raw_output = str(crew.kickoff()).strip()
730
 
731
+ # Restore original tool
732
+ doctor_news.tools = [original_gather]
733
+
734
+ # Parse JSON
735
+ result = _parse_json_output(raw_output)
736
+
737
+ # Validate structure
738
  for key in ["titles", "snippets", "urls"]:
739
  if key not in result or not isinstance(result[key], list):
740
+ raise ValueError(f"Missing or invalid '{key}' in result")
741
  if not (len(result["titles"]) == len(result["snippets"]) == len(result["urls"])):
742
+ raise ValueError("Mismatched list lengths in doctor news result")
743
 
744
  return {"status": "success", **result}
745
 
746
  except Exception as e:
747
  return {"status": "error", "message": str(e)}