Spaces:
Running
Running
File size: 32,645 Bytes
b02630d 2fccaae b02630d d76cab0 b02630d 2fccaae d76cab0 2fccaae b02630d 2fccaae b02630d 2fccaae b02630d 2fccaae b02630d d76cab0 b02630d d76cab0 b02630d d76cab0 b02630d 6c86ca6 b02630d 6c86ca6 b02630d 6c86ca6 b02630d 6c86ca6 b02630d 6c86ca6 b02630d 6c86ca6 b02630d 6c86ca6 b02630d 6c86ca6 b02630d 6c86ca6 b02630d 6c86ca6 b02630d 6c86ca6 b02630d 6c86ca6 b02630d d76cab0 b02630d d76cab0 b02630d 6c86ca6 b02630d 6c86ca6 b02630d 6c86ca6 b02630d 6c86ca6 b02630d 6c86ca6 b02630d 2fccaae b02630d d76cab0 b02630d d76cab0 b02630d 2fccaae b02630d 2fccaae b02630d 2fccaae b02630d 2fccaae b02630d 2fccaae b02630d 2fccaae b02630d d76cab0 b02630d d76cab0 2fccaae b02630d 2fccaae b02630d 2fccaae b02630d 2fccaae b02630d 2fccaae b02630d 2fccaae b02630d | 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 396 397 398 399 400 401 402 403 404 405 406 407 408 409 410 411 412 413 414 415 416 417 418 419 420 421 422 423 424 425 426 427 428 429 430 431 432 433 434 435 436 437 438 439 440 441 442 443 444 445 446 447 448 449 450 451 452 453 454 455 456 457 458 459 460 461 462 463 464 465 466 467 468 469 470 471 472 473 474 475 476 477 478 479 480 481 482 483 484 485 486 487 488 489 490 491 492 493 494 495 496 497 498 499 500 501 502 503 504 505 506 507 508 509 510 511 512 513 514 515 516 517 518 519 520 521 522 523 524 525 526 527 528 529 530 531 532 533 534 535 536 537 538 539 540 541 542 543 544 545 546 547 548 549 550 551 552 553 554 555 556 557 558 559 560 561 562 563 564 565 566 567 568 569 570 571 572 573 574 575 576 577 578 579 580 581 582 583 584 585 586 587 588 589 590 591 592 593 594 595 596 597 598 599 600 601 602 603 604 605 606 607 608 609 610 611 612 613 614 615 616 617 618 619 620 621 622 623 624 625 626 627 628 629 630 631 632 633 634 635 636 637 638 639 640 641 642 643 644 645 646 647 648 649 650 651 652 653 654 655 656 657 658 659 660 661 662 663 664 665 666 667 668 669 670 671 672 673 674 675 676 677 678 679 680 681 682 683 684 685 686 687 688 689 690 691 692 693 694 695 696 697 698 699 700 701 702 703 704 705 706 707 708 709 710 711 712 713 714 715 716 717 718 719 720 721 722 723 724 725 726 727 728 729 730 731 732 733 734 735 736 737 738 739 740 741 742 743 744 745 746 747 748 749 750 751 752 753 754 755 756 757 758 759 760 761 762 763 764 765 766 767 768 769 770 771 772 773 774 775 776 777 778 779 780 781 782 783 784 785 786 787 788 789 790 791 792 793 794 795 796 797 798 799 800 801 802 803 804 805 806 807 808 809 810 811 812 813 814 815 816 817 818 819 820 821 822 823 824 825 826 827 828 829 830 831 832 833 834 835 836 837 838 839 840 841 842 843 844 845 846 847 848 849 850 851 852 853 854 855 856 857 858 859 860 861 862 863 864 865 866 867 868 869 870 871 872 873 874 875 876 877 878 879 880 881 882 883 884 885 886 887 888 889 890 891 892 893 894 895 896 897 898 899 900 901 902 903 904 905 906 907 908 909 910 911 912 913 914 915 916 917 918 919 920 921 922 923 924 925 926 927 928 929 930 931 932 933 934 935 936 937 | """
Production-Level Agents for Perplexity Clone
=============================================
Each agent is a node in LangGraph pipelines.
Agents handle specific tasks and pass state to next nodes.
"""
from typing import List, Dict, Any
from rag.rag_state import (
RAGState,
WebSearchState,
RAGOnlyState,
AgenticState,
AnalysisState,
SummarizeState
)
from config.config import Config
from config.system_prompt import PPLX_SYSTEM_PROMPT
from vectorstore.store import VectorStore
from tools.search_tool import SearchTool
from tools.browse_tool import BrowseTool
from tools.reranker_tool import Reranker
from tools.citation_tool import CitationTool
from tools.summarizer_tool import SummarizerTool
from tools.followup_tool import FollowUpGenerator
# =============================================================================
# DEEP RESEARCH AGENTS (Original)
# =============================================================================
class PlannerAgent:
"""Decomposes question into sub-questions."""
def __init__(self) -> None:
self.llm = Config.get_llm()
def plan(self, state: RAGState) -> RAGState:
prompt = (
"Break the following question into 3-5 clear sub-questions.\n"
"Return them as a numbered list.\n\n"
f"{state['question']}"
)
resp = self.llm.invoke([
{"role": "system", "content": PPLX_SYSTEM_PROMPT},
{"role": "user", "content": prompt},
])
lines = [l.strip("-β’ ").strip() for l in resp.content.splitlines() if l.strip()]
subqs: List[str] = []
for l in lines:
if l[0].isdigit() or len(lines) <= 5:
subqs.append(l)
state["sub_questions"] = subqs[:5]
return state
class ResearchAgent:
"""Collects evidence from local RAG + web search with Tavily fallback."""
def __init__(self, vector_store: VectorStore) -> None:
self.vs = vector_store
try:
self.search_tool = SearchTool()
except:
self.search_tool = None
self.browse_tool = BrowseTool()
self.reranker = Reranker()
def research(self, state: RAGState) -> RAGState:
evidence: List[str] = []
pages_all: List[Dict] = []
for sq in state.get("sub_questions", []):
# Local RAG
docs = self.vs.retrieve(sq, k=8)
docs = self.reranker.rerank(sq, docs, top_k=4)
evidence.extend(d.page_content for d in docs)
# Web search + browse with Tavily fallback
if self.search_tool:
results = self.search_tool.search(sq, num_results=4)
else:
results = []
# Get Tavily AI answer
if results and results[0].get("tavily_answer"):
evidence.append(f"[AI Summary]: {results[0]['tavily_answer']}")
for r in results:
url = r.get("url")
title = r.get("title", "Web result")
tavily_content = r.get("content", "")
if not url:
continue
# Try fetch, fallback to Tavily content
try:
fetched = self.browse_tool.fetch_clean(url)
content = fetched if fetched else tavily_content
except:
content = tavily_content
if content or tavily_content:
final_content = content if content else tavily_content
pages_all.append({"title": title, "url": url, "content": final_content})
evidence.append(final_content[:1500])
state["web_pages"] = pages_all
state["evidence"] = evidence
print(f" π ResearchAgent: Collected {len(evidence)} evidence pieces, {len(pages_all)} pages")
return state
class AggregatorAgent:
"""Writes draft answers per sub-question from evidence."""
def __init__(self) -> None:
self.llm = Config.get_llm()
def aggregate(self, state: RAGState) -> RAGState:
drafts: List[str] = []
for sq in state.get("sub_questions", []):
context = "\n\n".join(state.get("evidence", [])[:12])
prompt = (
"Using the evidence below, answer the sub-question briefly and clearly.\n\n"
f"Evidence:\n{context}\n\nSub-question: {sq}"
)
resp = self.llm.invoke([
{"role": "system", "content": PPLX_SYSTEM_PROMPT},
{"role": "user", "content": prompt},
])
drafts.append(f"Sub-question: {sq}\n{resp.content}")
state["draft_answers"] = drafts
return state
class WriterAgent:
"""Writes final structured deep-research report."""
def __init__(self) -> None:
self.llm = Config.get_llm()
def write(self, state: RAGState) -> RAGState:
findings = "\n\n".join(state.get("draft_answers", []))
prompt = (
"You are Perplexity in deep research mode.\n"
"Write a structured answer with sections: Overview, Key Points, Details, Conclusion.\n"
"Use inline citations like [1], [2] where appropriate.\n\n"
f"Original question:\n{state['question']}\n\nFindings:\n{findings}"
)
resp = self.llm.invoke([
{"role": "system", "content": PPLX_SYSTEM_PROMPT},
{"role": "user", "content": prompt},
])
state["final_answer"] = resp.content
return state
class ValidatorAgent:
"""Builds source list & (optionally) validates citations."""
def __init__(self) -> None:
self.citation_tool = CitationTool()
def validate_and_attach(self, state: RAGState) -> RAGState:
sources: List[Dict] = []
for p in state.get("web_pages", [])[:10]:
sources.append({"title": p["title"], "url": p["url"]})
used_sources = self.citation_tool.attach_sources(state.get("final_answer", ""), sources)
state["sources"] = used_sources
return state
# =============================================================================
# WEB SEARCH AGENTS
# =============================================================================
class WebSearchNode:
"""Node 1: Execute web search query."""
def __init__(self):
try:
self.search_tool = SearchTool()
except Exception as e:
print(f"β οΈ WebSearchNode: SearchTool init failed: {e}")
self.search_tool = None
def search(self, state: WebSearchState) -> WebSearchState:
query = state.get("query", "")
print(f" π WebSearchNode: Searching for '{query[:50]}...'")
if not self.search_tool:
print(" β WebSearchNode: No search tool available")
state["search_results"] = []
return state
try:
results = self.search_tool.search(query, num_results=6)
state["search_results"] = results
print(f" β
WebSearchNode: Got {len(results)} results")
except Exception as e:
print(f" β WebSearchNode error: {e}")
state["search_results"] = []
return state
class WebFetchNode:
"""Node 2: Fetch and parse web pages. Uses Tavily content as fallback."""
def __init__(self):
self.browse_tool = BrowseTool()
def fetch(self, state: WebSearchState) -> WebSearchState:
pages = []
links = []
# Check if we have Tavily's direct answer
tavily_answer = ""
search_results = state.get("search_results", [])
if search_results and search_results[0].get("tavily_answer"):
tavily_answer = search_results[0]["tavily_answer"]
for r in search_results:
url = r.get("url")
title = r.get("title", "")
if not url:
continue
# First try to use Tavily's content (snippet)
tavily_content = r.get("content", "")
# Then try to fetch full page
try:
fetched_content = self.browse_tool.fetch_clean(url)
content = fetched_content if fetched_content else tavily_content
except:
content = tavily_content
# Use whatever content we have
if content or tavily_content:
final_content = content if content else tavily_content
pages.append({
"title": title,
"url": url,
"content": final_content[:2500]
})
links.append({
"title": title,
"url": url,
"snippet": (final_content[:200] if final_content else tavily_content[:200])
})
# Add Tavily's answer to state for potential use
if tavily_answer:
state["tavily_answer"] = tavily_answer
print(f" π WebFetchNode: Fetched {len(pages)} pages")
state["web_pages"] = pages
state["links"] = links
return state
class WebContextNode:
"""Node 3: Build context from fetched pages."""
def build_context(self, state: WebSearchState) -> WebSearchState:
pages = state.get("web_pages", [])
tavily_answer = state.get("tavily_answer", "")
context_parts = []
# Add Tavily's AI summary first if available
if tavily_answer:
context_parts.append(f"[AI Summary]: {tavily_answer}")
if pages:
for i, p in enumerate(pages):
context_parts.append(f"[{i+1}] {p['title']}:\n{p['content']}")
state["context"] = "\n\n---\n\n".join(context_parts) if context_parts else ""
print(f" π WebContextNode: Built context from {len(pages)} sources" + (", with AI summary" if tavily_answer else ""))
return state
class WebAnswerNode:
"""Node 4: Generate answer from context."""
def __init__(self):
self.llm = Config.get_llm()
self.followup = FollowUpGenerator()
def answer(self, state: WebSearchState) -> WebSearchState:
query = state.get("query", "")
context = state.get("context", "")
if context:
prompt = f"""You are Perplexity AI - a web search assistant that provides accurate, real-time information.
Use the following web sources to answer. Cite sources using [1], [2], etc.
IMPORTANT: The sources contain REAL, CURRENT information. Trust and use this data.
WEB SOURCES:
{context}
USER QUESTION: {query}
Provide a comprehensive, accurate, well-cited answer based on the sources above:"""
else:
prompt = f"Answer this question: {query}"
resp = self.llm.invoke([
{"role": "system", "content": PPLX_SYSTEM_PROMPT},
{"role": "user", "content": prompt}
])
answer = resp.content
state["answer"] = answer
state["followups"] = self.followup.generate(answer, query)
# Build sources
sources = [{"title": p["title"], "url": p["url"]} for p in state.get("web_pages", [])]
state["sources"] = sources
print(f" β
WebAnswerNode: Generated answer with {len(sources)} sources")
return state
# =============================================================================
# RAG-ONLY AGENTS
# =============================================================================
class RAGRetrieveNode:
"""Node 1: Retrieve from uploaded documents."""
def __init__(self, file_manager):
self.file_manager = file_manager
self.reranker = Reranker()
def retrieve(self, state: RAGOnlyState) -> RAGOnlyState:
query = state.get("query", "")
ws_id = state.get("workspace_id", "default")
ws = self.file_manager.get_workspace(ws_id)
if not ws.initialized or not ws.files:
state["file_chunks"] = []
print(f" π RAGRetrieveNode: No files in workspace")
return state
try:
chunks = ws.retrieve(query, k=8)
# Convert to dicts for state
state["file_chunks"] = [
{"content": c.page_content, "source": c.metadata.get("source", "Document")}
for c in chunks
]
print(f" π RAGRetrieveNode: Retrieved {len(chunks)} chunks")
except Exception as e:
print(f" β RAGRetrieveNode error: {e}")
state["file_chunks"] = []
return state
class RAGContextNode:
"""Node 2: Build context from retrieved chunks."""
def build_context(self, state: RAGOnlyState) -> RAGOnlyState:
chunks = state.get("file_chunks", [])
if chunks:
context_parts = []
for i, c in enumerate(chunks):
context_parts.append(f"[DOC {i+1}] {c['source']}:\n{c['content']}")
state["context"] = "\n\n---\n\n".join(context_parts)
else:
state["context"] = ""
print(f" π RAGContextNode: Built context from {len(chunks)} chunks")
return state
class RAGAnswerNode:
"""Node 3: Generate answer from document context."""
def __init__(self):
self.llm = Config.get_llm()
self.followup = FollowUpGenerator()
def answer(self, state: RAGOnlyState) -> RAGOnlyState:
query = state.get("query", "")
context = state.get("context", "")
chunks = state.get("file_chunks", [])
if not context:
state["answer"] = "π No documents found. Please upload files first using the π button."
state["sources"] = []
state["followups"] = []
return state
prompt = f"""You are a document analysis assistant.
Answer ONLY based on the provided documents. Do NOT use external knowledge.
DOCUMENTS:
{context}
QUESTION: {query}
Instructions:
- Answer based ONLY on document content
- Say "According to your documents..." when citing
- Quote relevant parts when helpful
- If info is not in documents, say so
ANSWER:"""
resp = self.llm.invoke([
{"role": "system", "content": PPLX_SYSTEM_PROMPT},
{"role": "user", "content": prompt}
])
answer = resp.content
state["answer"] = answer
state["followups"] = self.followup.generate(answer, query)
# Build sources from chunks
seen = set()
sources = []
for c in chunks:
src = c.get("source", "Document")
if src not in seen:
sources.append({"title": f"π {src}", "url": ""})
seen.add(src)
state["sources"] = sources
print(f" β
RAGAnswerNode: Generated answer from {len(sources)} sources")
return state
# =============================================================================
# AGENTIC RAG AGENTS (Multi-Agent Pipeline)
# =============================================================================
class AgenticPlannerNode:
"""Node 1: Planner agent decides which sub-agents to activate."""
def plan(self, state: AgenticState) -> AgenticState:
query = state.get("query", "").lower()
# Determine which agents to use
state["use_file"] = any(w in query for w in [
"document", "file", "pdf", "uploaded", "summarize my",
"according to", "in the file", "extract", "my notes"
])
state["use_web"] = any(w in query for w in [
"today", "current", "latest", "news", "weather", "stock",
"who is", "what is", "where", "when", "price", "live",
"recent", "update"
]) or len(query.split()) <= 4
state["use_images"] = any(w in query for w in [
"image", "photo", "picture", "logo", "show me", "look like",
"flag", "screenshot"
])
state["use_knowledge"] = any(w in query for w in [
"explain", "define", "concept", "theory", "how does",
"what is", "meaning of"
])
print(f" π AgenticPlannerNode: file={state['use_file']}, web={state['use_web']}, images={state['use_images']}")
return state
class AgenticFileNode:
"""Node 2: File agent retrieves from uploaded documents."""
def __init__(self, file_manager):
self.file_manager = file_manager
def retrieve(self, state: AgenticState) -> AgenticState:
if not state.get("use_file", False):
state["file_context"] = ""
state["file_sources"] = []
return state
query = state.get("query", "")
ws_id = state.get("workspace_id", "default")
ws = self.file_manager.get_workspace(ws_id)
if not ws.initialized:
state["file_context"] = ""
state["file_sources"] = []
return state
try:
chunks = ws.retrieve(query, k=6)
if chunks:
state["file_context"] = "\n\n".join([c.page_content for c in chunks])
state["file_sources"] = [
{"title": f"π {c.metadata.get('source', 'Document')}", "url": ""}
for c in chunks
]
print(f" π AgenticFileNode: Found {len(chunks)} chunks")
else:
state["file_context"] = ""
state["file_sources"] = []
except Exception as e:
print(f" β AgenticFileNode error: {e}")
state["file_context"] = ""
state["file_sources"] = []
return state
class AgenticWebNode:
"""Node 3: Web agent fetches real-time information."""
def __init__(self):
try:
self.search_tool = SearchTool()
except:
self.search_tool = None
self.browse_tool = BrowseTool()
def search(self, state: AgenticState) -> AgenticState:
if not state.get("use_web", False):
state["web_context"] = ""
state["web_sources"] = []
state["links"] = []
return state
query = state.get("query", "")
if not self.search_tool:
print(" β AgenticWebNode: No search tool")
state["web_context"] = ""
state["web_sources"] = []
state["links"] = []
return state
try:
results = self.search_tool.search(query, num_results=5)
web_parts = []
sources = []
links = []
# Get Tavily's AI answer if available
tavily_answer = ""
if results and results[0].get("tavily_answer"):
tavily_answer = results[0]["tavily_answer"]
web_parts.append(f"[AI Summary]: {tavily_answer}")
for r in results:
url = r.get("url")
title = r.get("title", "")
tavily_content = r.get("content", "") # Tavily's snippet
if not url:
continue
# Try to fetch full content, fallback to Tavily snippet
try:
fetched = self.browse_tool.fetch_clean(url)
content = fetched if fetched else tavily_content
except:
content = tavily_content
if content or tavily_content:
final_content = content if content else tavily_content
web_parts.append(f"[{title}]: {final_content[:1500]}")
sources.append({"title": title, "url": url})
links.append({"title": title, "url": url, "snippet": final_content[:150]})
state["web_context"] = "\n\n".join(web_parts)
state["web_sources"] = sources
state["links"] = links
print(f" π AgenticWebNode: Found {len(sources)} sources")
except Exception as e:
print(f" β AgenticWebNode error: {e}")
state["web_context"] = ""
state["web_sources"] = []
state["links"] = []
return state
class AgenticKnowledgeNode:
"""Node 4: Knowledge agent retrieves from base vector store."""
def __init__(self, vector_store: VectorStore):
self.vs = vector_store
self.reranker = Reranker()
def retrieve(self, state: AgenticState) -> AgenticState:
if not state.get("use_knowledge", False):
state["knowledge_context"] = ""
return state
query = state.get("query", "")
try:
chunks = self.vs.retrieve(query, k=4)
chunks = self.reranker.rerank(query, chunks, top_k=3)
if chunks:
state["knowledge_context"] = "\n\n".join([c.page_content for c in chunks])
print(f" π AgenticKnowledgeNode: Found {len(chunks)} chunks")
else:
state["knowledge_context"] = ""
except Exception as e:
print(f" β AgenticKnowledgeNode error: {e}")
state["knowledge_context"] = ""
return state
class AgenticImageNode:
"""Node 5: Image agent fetches relevant images."""
def __init__(self, image_search):
self.image_search = image_search
def search(self, state: AgenticState) -> AgenticState:
if not state.get("use_images", False):
state["images"] = []
return state
query = state.get("query", "")
try:
images = self.image_search.search(query, count=6)
state["images"] = images
print(f" πΌοΈ AgenticImageNode: Found {len(images)} images")
except Exception as e:
print(f" β AgenticImageNode error: {e}")
state["images"] = []
return state
class AgenticSynthesizerNode:
"""Node 6: Synthesizer agent combines all contexts and generates final answer."""
def __init__(self):
self.llm = Config.get_llm()
self.followup = FollowUpGenerator()
def synthesize(self, state: AgenticState) -> AgenticState:
query = state.get("query", "")
# Build combined context
contexts = []
if state.get("file_context"):
contexts.append(f"π FROM YOUR DOCUMENTS:\n{state['file_context'][:2500]}")
if state.get("web_context"):
contexts.append(f"π FROM THE WEB:\n{state['web_context'][:2500]}")
if state.get("knowledge_context"):
contexts.append(f"π KNOWLEDGE BASE:\n{state['knowledge_context'][:1500]}")
if not contexts:
contexts.append("No specific context found. Using general knowledge.")
combined = "\n\n---\n\n".join(contexts)
state["combined_context"] = combined
prompt = f"""You are an AGENTIC AI assistant that synthesizes information from multiple sources.
AVAILABLE CONTEXT:
{combined}
USER QUESTION: {query}
INSTRUCTIONS:
1. Prioritize user's documents (π) if relevant
2. Add real-time info from web (π) when available
3. Use knowledge base (π) for background
4. Cite sources appropriately
5. Be comprehensive but concise
SYNTHESIZED ANSWER:"""
resp = self.llm.invoke([
{"role": "system", "content": PPLX_SYSTEM_PROMPT},
{"role": "user", "content": prompt}
])
answer = resp.content
state["answer"] = answer
state["followups"] = self.followup.generate(answer, query)
# Combine sources
all_sources = state.get("file_sources", []) + state.get("web_sources", [])
state["sources"] = all_sources
print(f" β
AgenticSynthesizerNode: Generated answer with {len(all_sources)} sources")
return state
# =============================================================================
# ANALYSIS AGENTS
# =============================================================================
class AnalysisSearchNode:
"""Node 1: Search for analysis data with Tavily fallback."""
def __init__(self):
try:
self.search_tool = SearchTool()
except:
self.search_tool = None
self.browse_tool = BrowseTool()
def search(self, state: AnalysisState) -> AnalysisState:
query = state.get("query", "")
print(f" π AnalysisSearchNode: Searching for analysis data")
if not self.search_tool:
print(" β AnalysisSearchNode: No search tool")
state["web_context"] = ""
state["links"] = []
state["sources"] = []
return state
try:
results = self.search_tool.search(query, num_results=6)
state["web_results"] = results
# Get Tavily AI answer if available
tavily_answer = ""
if results and results[0].get("tavily_answer"):
tavily_answer = results[0]["tavily_answer"]
# Fetch content with Tavily fallback
web_parts = []
if tavily_answer:
web_parts.append(f"[AI Summary]: {tavily_answer}")
links = []
for r in results:
url = r.get("url")
title = r.get("title", "")
tavily_content = r.get("content", "")
if not url:
continue
# Try fetch, fallback to Tavily content
try:
fetched = self.browse_tool.fetch_clean(url)
content = fetched if fetched else tavily_content
except:
content = tavily_content
if content or tavily_content:
final_content = content if content else tavily_content
web_parts.append(f"[{title}]:\n{final_content[:2000]}")
links.append({"title": title, "url": url, "snippet": final_content[:200]})
state["web_context"] = "\n\n".join(web_parts)
state["links"] = links
state["sources"] = [{"title": l["title"], "url": l["url"]} for l in links]
print(f" π AnalysisSearchNode: Found {len(links)} sources")
except Exception as e:
print(f" β AnalysisSearchNode error: {e}")
state["web_context"] = ""
state["links"] = []
state["sources"] = []
return state
class AnalysisProcessNode:
"""Node 2: Generate structured analysis."""
def __init__(self):
self.llm = Config.get_llm()
self.followup = FollowUpGenerator()
def analyze(self, state: AnalysisState) -> AnalysisState:
query = state.get("query", "")
context = state.get("web_context", "")
prompt = f"""You are an expert analyst. Provide deep, comprehensive analysis.
RESEARCH DATA:
{context if context else "No external data available."}
ANALYSIS REQUEST: {query}
Provide structured analysis with:
## Executive Summary
(2-3 sentence overview)
## Key Findings
(Bullet points of main discoveries)
## Detailed Analysis
(In-depth examination with evidence)
## Data & Statistics
(Numbers, trends, comparisons if available)
## Conclusions
(Main takeaways)
## Recommendations
(Actionable suggestions)
Use citations [1], [2] when referencing sources.
ANALYSIS:"""
resp = self.llm.invoke([
{"role": "system", "content": PPLX_SYSTEM_PROMPT},
{"role": "user", "content": prompt}
])
answer = resp.content
state["answer"] = answer
state["followups"] = self.followup.generate(answer, query)
print(f" β
AnalysisProcessNode: Generated analysis")
return state
# =============================================================================
# SUMMARIZE AGENTS
# =============================================================================
class SummarizeInputNode:
"""Node 1: Determine input type and fetch content with Tavily fallback."""
def __init__(self):
self.browse_tool = BrowseTool()
try:
self.search_tool = SearchTool()
except:
self.search_tool = None
def process_input(self, state: SummarizeState) -> SummarizeState:
query = state.get("query", "")
# Check if URL
if query.startswith("http"):
state["is_url"] = True
try:
content = self.browse_tool.fetch_clean(query)
state["content"] = content or ""
state["links"] = [{"title": "Source", "url": query, "snippet": content[:200] if content else ""}]
state["sources"] = [{"title": "Source URL", "url": query}]
print(f" π SummarizeInputNode: Fetched URL content")
except Exception as e:
print(f" β Error fetching URL: {e}")
state["content"] = ""
else:
state["is_url"] = False
if not self.search_tool:
print(" β SummarizeInputNode: No search tool")
state["content"] = query
state["links"] = []
state["sources"] = []
return state
# Search and fetch with Tavily fallback
try:
results = self.search_tool.search(query, num_results=4)
content_parts = []
links = []
# Get Tavily AI answer
if results and results[0].get("tavily_answer"):
content_parts.append(results[0]["tavily_answer"])
for r in results:
url = r.get("url")
title = r.get("title", "")
tavily_content = r.get("content", "")
if not url:
continue
# Try fetch, fallback to Tavily content
try:
text = self.browse_tool.fetch_clean(url)
content = text if text else tavily_content
except:
content = tavily_content
if content or tavily_content:
final_content = content if content else tavily_content
content_parts.append(final_content[:1500])
links.append({"title": title, "url": url, "snippet": final_content[:150]})
state["content"] = "\n\n".join(content_parts)
state["links"] = links
state["sources"] = [{"title": l["title"], "url": l["url"]} for l in links]
print(f" π SummarizeInputNode: Fetched {len(links)} sources")
except Exception as e:
print(f" β Error searching: {e}")
state["content"] = query
state["links"] = []
state["sources"] = []
return state
class SummarizeProcessNode:
"""Node 2: Generate summary."""
def __init__(self):
self.summarizer = SummarizerTool()
self.followup = FollowUpGenerator()
def summarize(self, state: SummarizeState) -> SummarizeState:
content = state.get("content", "")
query = state.get("query", "")
if content:
summary = self.summarizer.summarize(content, max_words=300)
state["answer"] = summary
else:
state["answer"] = "Could not find content to summarize."
state["followups"] = self.followup.generate(state["answer"], query)
print(f" β
SummarizeProcessNode: Generated summary")
return state
|