Spaces:
Sleeping
Sleeping
File size: 48,635 Bytes
a00a15f 81b8302 a00a15f 81b8302 a00a15f 384d517 a00a15f 384d517 a00a15f 3cf6d28 a00a15f 384d517 a00a15f 384d517 a00a15f 384d517 a00a15f 384d517 a00a15f 384d517 a00a15f 81b8302 a00a15f 81b8302 a00a15f 384d517 a00a15f 81b8302 a00a15f 81b8302 384d517 a00a15f 81b8302 a00a15f 81b8302 a00a15f 81b8302 a00a15f 81b8302 a00a15f 81b8302 a00a15f 81b8302 a00a15f 81b8302 a00a15f 81b8302 a00a15f 81b8302 a00a15f 81b8302 a00a15f 81b8302 a00a15f 81b8302 a00a15f 81b8302 a00a15f 81b8302 a00a15f 81b8302 a00a15f 81b8302 a00a15f 81b8302 a00a15f 81b8302 a00a15f 384d517 a00a15f 384d517 a00a15f 384d517 a00a15f 384d517 a00a15f 384d517 a00a15f 384d517 a00a15f 384d517 a00a15f 81b8302 a00a15f 81b8302 a00a15f 81b8302 384d517 a00a15f 81b8302 a00a15f 81b8302 a00a15f 384d517 81b8302 b34c33b a00a15f 81b8302 384d517 b34c33b 384d517 a00a15f 384d517 a00a15f 384d517 a00a15f 384d517 81b8302 384d517 a00a15f 384d517 a00a15f 384d517 a00a15f 81b8302 a00a15f 81b8302 384d517 81b8302 384d517 81b8302 384d517 81b8302 384d517 a00a15f 81b8302 384d517 81b8302 a00a15f 384d517 a00a15f 81b8302 384d517 a00a15f 81b8302 a00a15f 81b8302 a00a15f 81b8302 a00a15f 384d517 81b8302 a00a15f 384d517 a00a15f 81b8302 384d517 a00a15f 81b8302 a00a15f 81b8302 a00a15f 81b8302 384d517 81b8302 a00a15f 384d517 81b8302 a00a15f 81b8302 a00a15f 81b8302 a00a15f 81b8302 a00a15f 81b8302 a00a15f 81b8302 a00a15f 384d517 a00a15f 384d517 a00a15f 384d517 a00a15f 384d517 a00a15f 384d517 a00a15f 384d517 a00a15f 384d517 a00a15f 81b8302 a00a15f 81b8302 a00a15f 384d517 81b8302 a00a15f 384d517 a00a15f 384d517 a00a15f 81b8302 a00a15f 384d517 a00a15f 81b8302 a00a15f 81b8302 a00a15f 59ceb93 a00a15f 81b8302 a00a15f 81b8302 a00a15f 384d517 a00a15f 384d517 81b8302 a00a15f 81b8302 a00a15f 81b8302 a00a15f 81b8302 a00a15f 384d517 a00a15f 81b8302 384d517 81b8302 a00a15f 384d517 a00a15f 81b8302 a00a15f 81b8302 384d517 81b8302 a00a15f 384d517 a00a15f 384d517 a00a15f 384d517 a00a15f f21f2d1 a00a15f 81b8302 | 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 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 396 397 398 399 400 401 402 403 404 405 406 407 408 409 410 411 412 413 414 415 416 417 418 419 420 421 422 423 424 425 426 427 428 429 430 431 432 433 434 435 436 437 438 439 440 441 442 443 444 445 446 447 448 449 450 451 452 453 454 455 456 457 458 459 460 461 462 463 464 465 466 467 468 469 470 471 472 473 474 475 476 477 478 479 480 481 482 483 484 485 486 487 488 489 490 491 492 493 494 495 496 497 498 499 500 501 502 503 504 505 506 507 508 509 510 511 512 513 514 515 516 517 518 519 520 521 522 523 524 525 526 527 528 529 530 531 532 533 534 535 536 537 538 539 540 541 542 543 544 545 546 547 548 549 550 551 552 553 554 555 556 557 558 559 560 561 562 563 564 565 566 567 568 569 570 571 572 573 574 575 576 577 578 579 580 581 582 583 584 585 586 587 588 589 590 591 592 593 594 595 596 597 598 599 600 601 602 603 604 605 606 607 608 609 610 611 612 613 614 615 616 617 618 619 620 621 622 623 624 625 626 627 628 629 630 631 632 633 634 635 636 637 638 639 640 641 642 643 644 645 646 647 648 649 650 651 652 653 654 655 656 657 658 659 660 661 662 663 664 665 666 667 668 669 670 671 672 673 674 675 676 677 678 679 680 681 682 683 684 685 686 687 688 689 690 691 692 693 694 695 696 697 698 699 700 701 702 703 704 705 706 707 708 709 710 711 712 713 714 715 716 717 718 719 720 721 722 723 724 725 726 727 728 729 730 731 732 733 734 735 736 737 738 739 740 741 742 743 744 745 746 747 748 749 750 751 752 753 754 755 756 757 758 759 760 761 762 763 764 765 766 767 768 769 770 771 772 773 774 775 776 777 778 779 780 781 782 783 784 785 786 787 788 789 790 791 792 793 794 795 796 797 798 799 800 801 802 803 804 805 806 807 808 809 810 811 812 813 814 815 816 817 818 819 820 821 822 823 824 825 826 827 828 829 830 831 832 833 834 835 836 837 838 839 840 841 842 843 844 845 846 847 848 849 850 851 852 853 854 855 856 857 858 859 860 861 862 863 864 865 866 867 868 869 870 871 872 873 874 875 876 877 878 879 880 881 882 883 884 885 886 887 888 889 890 891 892 893 894 895 896 897 898 899 900 901 902 903 904 905 906 907 908 909 910 911 912 913 914 915 916 917 918 919 920 921 922 923 924 925 926 927 928 929 930 931 932 933 934 935 936 937 938 939 940 941 942 943 944 945 946 947 948 949 950 951 952 953 954 955 956 957 958 959 960 961 962 963 964 965 966 967 968 969 970 971 972 973 974 975 976 977 978 979 980 981 982 983 984 985 986 987 988 989 990 991 992 993 994 995 996 997 998 999 1000 1001 1002 1003 1004 1005 1006 1007 1008 1009 1010 1011 1012 1013 1014 1015 1016 1017 1018 1019 1020 1021 1022 1023 1024 1025 1026 1027 1028 1029 1030 1031 1032 1033 1034 1035 1036 1037 1038 1039 1040 1041 1042 1043 1044 1045 1046 1047 1048 1049 1050 1051 1052 1053 1054 1055 1056 1057 1058 1059 1060 1061 1062 1063 1064 1065 1066 1067 1068 1069 1070 1071 1072 1073 1074 1075 1076 1077 1078 1079 1080 1081 1082 1083 1084 1085 1086 1087 1088 1089 1090 1091 1092 1093 1094 1095 1096 1097 1098 1099 1100 1101 1102 1103 1104 1105 1106 1107 1108 1109 1110 1111 1112 1113 1114 1115 1116 1117 1118 1119 1120 1121 1122 1123 1124 1125 1126 1127 1128 1129 1130 1131 1132 1133 1134 1135 1136 1137 1138 1139 1140 1141 1142 1143 1144 1145 1146 1147 1148 1149 1150 1151 1152 1153 1154 1155 1156 1157 1158 1159 1160 1161 1162 1163 1164 1165 1166 1167 1168 1169 1170 1171 1172 1173 | # %% [markdown]
# ## Academic Research Assistant
# %% [markdown]
# ## Import
# %%
from docling.document_converter import DocumentConverter
import tqdm as notebook_tqdm
from pydantic import BaseModel, Field
import os
from typing import Optional, Any, Literal, Dict, List, Tuple, Type, Annotated
from operator import add
from typing_extensions import TypedDict
from langgraph.graph import StateGraph, START, END
from langgraph.types import Command
from langchain_openai import ChatOpenAI
from langchain_google_genai import ChatGoogleGenerativeAI
from langchain_core.prompts import ChatPromptTemplate
from langchain_core.output_parsers import PydanticOutputParser # Add this import
# from langfuse.callback import CallbackHandler
import gradio as gr
import contextlib
from io import StringIO
import docx
from pathlib import Path
import re
from typing import Union
from dotenv import load_dotenv
# Load environment variables from .env file
load_dotenv()
# Use environment variables for API keys
USE_GOOGLE = False
API_KEY = os.environ.get("NEBIUS_KEY")
MODEL_NAME = None
ENDPOINT_URL = None
# Try these models one by one to see which ones actually exist
NEBIUS_MODELS = [
"meta-llama/Llama-2-7b-chat-hf", # Try this first
"mistralai/Mistral-7B-Instruct-v0.2", # Then this
"microsoft/DialoGPT-medium", # Then this
"openai/gpt-3.5-turbo", # Or this
"Qwen2.5-Coder-7B", # Keep the original as fallback
"QwQ-32B"
]
def list_nebius_models():
"""List all available models from Nebius API."""
try:
import requests
headers = {
"Authorization": f"Bearer {API_KEY}",
"Content-Type": "application/json"
}
# Try the models endpoint
response = requests.get(
f"{ENDPOINT_URL}models",
headers=headers,
timeout=10
)
if response.status_code == 200:
models = response.json()
print("Available models:")
for model in models.get('data', []):
print(f" - {model.get('id', 'Unknown')}")
return [model.get('id') for model in models.get('data', [])]
else:
print(f"Failed to fetch models: {response.status_code}")
print(f"Response: {response.text}")
return []
except Exception as e:
print(f"Error fetching models: {str(e)}")
return []
def test_available_models():
"""Test which models are actually available."""
# First try to get the actual model list
available_models = list_nebius_models()
if available_models:
print(f"Found {len(available_models)} models from API")
test_models = available_models[:6] # Test first 6 models
else:
# Fallback to common model names that might work
test_models = [
"gpt-3.5-turbo",
"gpt-4",
"claude-3-haiku",
"llama-2-7b-chat",
"mistral-7b-instruct",
"qwen-7b-chat"
]
for model in test_models:
try:
print(f"Testing model: {model}")
global MODEL_NAME
MODEL_NAME = model
# Simple test call
llm = ChatOpenAI(
model=model,
api_key=API_KEY,
base_url=ENDPOINT_URL,
max_completion_tokens=50,
timeout=10,
temperature=0
)
response = llm.invoke("Hello")
print(f"β
{model} works!")
return model # Return the first working model
except Exception as e:
print(f"β {model} failed: {str(e)}")
continue
print("β οΈ No working models found")
return None
# Call this function when setting up API key
def setup_api_key(nebius_key=None, model_name=None):
global API_KEY, MODEL_NAME, ENDPOINT_URL, USE_GOOGLE
# First try user-provided key (from UI)
if nebius_key:
API_KEY = nebius_key
ENDPOINT_URL = "https://api.studio.nebius.com/v1/"
# Test which model actually works
if model_name:
MODEL_NAME = model_name
else:
working_model = test_available_models()
if working_model:
MODEL_NAME = working_model
else:
print("No working models found")
return False
print(f"Using user-provided Nebius API key with model: {MODEL_NAME}")
return True
# Next try environment variable
if API_KEY:
ENDPOINT_URL = "https://api.studio.nebius.com/v1/"
if model_name:
MODEL_NAME = model_name
else:
working_model = test_available_models()
if working_model:
MODEL_NAME = working_model
else:
print("No working models found")
return False
print(f"Using Nebius API from environment variable with model: {MODEL_NAME}")
return True
print("No API key found. Please provide a Nebius API key.")
return False
# Initialize with environment variables if available
setup_api_key()
# %% [markdown]
# ## Structured outputs
# %%
class ResearchSummary(BaseModel):
key_findings: List[str] = Field(..., description="A list of the most important findings from the research paper.")
methodology: str = Field(..., description="A brief description of the methodology used in the research.")
limitations: List[str] = Field(..., description="A list of the limitations of the study as identified by the authors or the agent.")
class FutureScope(BaseModel):
identified_gaps: List[str] = Field(..., description="List of identified research gaps based on the provided paper(s).")
suggested_directions: List[str] = Field(..., description="Concrete suggestions for future research directions or next studies.")
synthesis: str = Field(..., description="A brief synthesis of how these future directions build upon the provided literature.")
class MultiStepPlan(BaseModel):
reasoning : str = Field("", description="The multi-step reasoning required to break down the user query in a plan.")
plan : List[Literal["summary_agent", "synthesis_agent", "future_scope_agent", "critique_agent"]] = Field("END", description="The list of agents required to fulfill the user request determined by the Orchestrator.")
class PaperSummary(BaseModel):
key_findings: List[str] = Field(
default_factory=lambda: ["No key findings available due to processing error"],
description="List of key findings from the paper"
)
methodology: str = Field(
default="Methodology not available due to processing error",
description="Summary of the methodology used in the paper"
)
conclusion: str = Field(
default="Conclusion not available due to processing error",
description="Summary of the paper's conclusion"
)
# %% [markdown]
# ## Agent state
# %%
class AgentDescription(TypedDict):
"Agent description containing the title, system prompt and description."
title : str
description : str
system_prompt : str
class ResearchAgentState(BaseModel):
"""State for the research agent."""
research_papers: Annotated[List[Tuple[str, str]], add] = Field(default_factory=list) # List of (filename, content) tuples
summary: Annotated[List[Dict], add] = Field(default_factory=list) # List of paper summaries
user_query: str = Field(default="") # Remove annotation - only set once
phase: str = Field(default="PLAN") # PLAN, EXECUTE, ANSWER
plan: List[str] = Field(default_factory=list) # List of agent names to call in order
messages: Annotated[List[Tuple[str, str]], add] = Field(default_factory=list) # List of (agent, message) tuples
critique: Optional[str] = Field(default=None) # Optional critique of the analysis
available_agents: Dict[str, Dict] = Field(default_factory=dict) # Mapping of agent name to agent description
final_answer: Optional[str] = Field(default=None) # Final answer to the user's query
max_iterations: int = Field(default=1) # Maximum number of iterations for processing
synthesis_of_findings: Optional[str] = Field(default=None) # Remove Annotated - only set once
future_directions_report: Optional[str] = Field(default=None) # Remove Annotated - only set once
# %% [markdown]
# ## System prompts
# %%
general_prefix = """
You are part of a collaborative multi-agent system called the *Academic Research Assistant*.
This system consists of specialized agents working together to analyze, synthesize, and critique academic literature.
Each agent has a distinct role. You are encouraged to build upon the work of other agents to produce a comprehensive and insightful analysis.
"""
summary_prompt = """
You are a diligent research assistant. Your task is to read the provided research paper and extract the most critical information.
Focus on the following key areas:
1. **Key Findings:** What were the main results and conclusions of the study? List them as clear, concise bullet points.
2. **Methodology:** Briefly describe the methodology, including the techniques, dataset, and experiments conducted.
3. **Limitations:** Identify any limitations of the study that were mentioned by the authors.
Provide your output in a structured format. Do not add any interpretation; stick strictly to the information present in the paper.
"""
synthesis_prompt = """
You are a research analyst specializing in literature reviews. You have been provided with summaries from one or more research papers.
Your task is to synthesize this information into a cohesive narrative.
1. Identify common themes, findings, and methodologies across the papers.
2. Highlight any conflicting or divergent results.
3. Create a single, flowing text that summarizes the current state of research based *only* on the provided information.
Do not introduce outside knowledge. Your synthesis should serve as a high-level overview for someone trying to understand the field as defined by these papers.
"""
future_scope_prompt = """
You are an experienced academic advisor with a knack for identifying promising research avenues.
Based on the provided research summaries and synthesis, your goal is to propose a clear path for future work.
Follow these steps:
1. **Identify Research Gaps:** Based on the limitations and findings of the papers, what questions remain unanswered? What are the clear gaps in the current body of knowledge?
2. **Suggest Future Directions:** Propose 2-3 concrete, actionable research projects that could address these gaps. For each suggestion, briefly explain:
- The research question.
- A potential methodology.
- The expected contribution to the field.
3. **Write a Concluding Synthesis:** Briefly summarize why these future directions are a logical and important next step in this research area.
Your tone should be formal and academic. The suggestions must be directly inspired by the provided context.
"""
critique_prompt = """
You are a peer reviewer. Your task is to provide constructive feedback on the generated research analysis (synthesis and future scope).
Evaluate the analysis based on the following criteria:
- **Clarity and Cohesion:** Is the synthesis clear, well-structured, and easy to understand?
- **Logical Flow:** Do the suggested future directions logically follow from the identified gaps and the provided literature?
- **Actionability:** Are the future scope suggestions concrete and specific enough to be pursued?
- **Completeness:** Does the analysis seem to have missed any obvious connections or gaps present in the source material?
Provide brief, constructive feedback highlighting points of improvement.
Provide a quality flag:
- **EXCELLENT**: The analysis is clear, logical, and insightful.
- **NEEDS REVISION**: The analysis has flaws in logic, clarity, or completeness that need to be addressed.
"""
# %% [markdown]
# ### Available agents
# %%
summary_agent_description = AgentDescription(
title="summary_agent",
description="Summarizes the key findings, methodology, and limitations of a single research paper.",
system_prompt=general_prefix + summary_prompt
)
synthesis_agent_description = AgentDescription(
title="synthesis_agent",
description="Synthesizes information from multiple paper summaries into a cohesive literature review.",
system_prompt=general_prefix + synthesis_prompt
)
future_scope_agent_description = AgentDescription(
title="future_scope_agent",
description="Identifies research gaps and suggests concrete directions for future work based on the literature.",
system_prompt=general_prefix + future_scope_prompt
)
critique_agent_description = AgentDescription(
title="critique_agent",
description="Provides peer-review style feedback on the generated synthesis and future scope analysis.",
system_prompt=general_prefix + critique_prompt
)
available_agents = {
"summary_agent": summary_agent_description,
"synthesis_agent": synthesis_agent_description,
"future_scope_agent": future_scope_agent_description,
"critique_agent": critique_agent_description,
}
# %% [markdown]
# ## Utilities
# %% [markdown]
# ### General utilities
# %%
def read_file_content(file: Union[str, Path]) -> str:
file_path = Path(file)
suffix = file_path.suffix.lower()
if suffix == ".txt" or suffix == ".md":
return file_path.read_text(encoding="utf-8")
elif suffix == ".pdf":
converter = DocumentConverter()
result = converter.convert(file_path)
return result.document.export_to_markdown()
elif suffix == ".docx":
return "\n".join(p.text for p in docx.Document(file_path).paragraphs)
else:
return ""
# %% [markdown]
# ## LLM call
# %%
def call_llm(system_prompt, user_prompt, response_format=None):
"""Call LLM with system and user prompt, optionally parsing to a specific format"""
global API_KEY, MODEL_NAME, ENDPOINT_URL
if not API_KEY:
print("Error: API key is not set")
# Return a default instance for the response_format class
if response_format and hasattr(response_format, "__name__"):
try:
if response_format.__name__ == "MultiStepPlan":
return MultiStepPlan(
reasoning="Error occurred: API key not set",
plan=["summary_agent", "synthesis_agent", "future_scope_agent"]
)
elif response_format.__name__ == "PaperSummary":
return PaperSummary() # Uses default values from Field definitions
else:
# Generic attempt to create an instance with default values
return response_format()
except Exception as e:
print(f"Failed to create default instance: {str(e)}")
return None
try:
if USE_GOOGLE:
llm = ChatGoogleGenerativeAI(
model=MODEL_NAME,
google_api_key=API_KEY,
temperature=0
)
else:
llm = ChatOpenAI(
model=MODEL_NAME,
api_key=API_KEY,
base_url=ENDPOINT_URL,
max_completion_tokens=None,
timeout=60,
max_retries=2,
temperature=0
)
if response_format is not None:
llm = llm.with_structured_output(response_format)
prompt = ChatPromptTemplate.from_messages([
("system", "{system_prompt}"),
("user", "{user_prompt}")
])
chain = prompt | llm
print(f"Calling model: {MODEL_NAME}")
response = chain.invoke({
"system_prompt": system_prompt,
"user_prompt": user_prompt
})
return response
except Exception as e:
print(f"Error in call_llm: {str(e)}")
if hasattr(e, 'response') and hasattr(e.response, 'json'):
try:
error_details = e.response.json()
print(f"API Error details: {error_details}")
except:
pass
# Create a default response based on the response_format class
if response_format and hasattr(response_format, "__name__"):
try:
if response_format.__name__ == "MultiStepPlan":
return MultiStepPlan(
reasoning="Error occurred while calling the LLM API. Using default plan.",
plan=["summary_agent", "synthesis_agent", "future_scope_agent"]
)
elif response_format.__name__ == "PaperSummary":
return PaperSummary() # Uses default values from Field definitions
else:
# Generic attempt to create an instance with default values
return response_format()
except Exception as e:
print(f"Failed to create default instance: {str(e)}")
return None
def serialize_messages(messages : List[Tuple[str,str]]) -> str:
"Returns a formatted message history of previous messages"
return "\n" +"\n".join(f"**{role}:**\n{content}" for role, content in messages)
def strip_think_blocks(text: str) -> str:
return re.sub(r"<think>.*?</think>", "", text, flags=re.DOTALL)
# %% [markdown]
# ### Gradio utilities
# %%
# Handle different result types cleanly
def type_conversion(obj : Any, type):
"Return the object in a gradio compatible type"
if isinstance(obj, type):
result_dict = obj.model_dump()
elif isinstance(obj, Dict):
result_dict = obj
else:
# Handle possible dataclass or similar object
try:
result_dict = ResearchAgentState.model_validate(obj).model_dump()
except Exception as e:
print(f"Error converting output of type {type(obj)}")
return result_dict
# %% [markdown]
# ## Agents
# %% [markdown]
# ### Orchestrator agent
#
#
# %%
def orchestrator_agent(state: ResearchAgentState) -> Command:
"""Central orchestration logic to determine the next agent to call."""
if not state.research_papers:
return Command(
goto=END,
update={"final_answer": "### βοΈ The research assistant needs at least one research paper to begin.\n" \
"ππ½ Please upload one or more research papers in the 'π Research Materials' tab."}
)
if state.phase == "PLAN":
agent_descriptions = "\n".join([
f"**{agent.get('title')}**: {agent.get('description')}"
for name, agent in state.available_agents.items()
])
system_prompt = f"""You are an orchestrator for an academic research assistant. Your task is to create a plan to answer the user's query using a team of specialized agents.
**Agents:**
{agent_descriptions}
Based on the user's query, create a logical sequence of agents to call. For example, to find future scope, you should first summarize the papers, then synthesize them, and then call the future_scope_agent.
**IMPORTANT:** Always include the summary_agent as the first step when working with research papers. Every task requires proper paper summaries before analysis can begin.
"""
user_prompt = state.user_query
response = call_llm(system_prompt, user_prompt, MultiStepPlan)
# Handle None response by providing a default plan
if response is None:
print("β οΈ Failed to get response from LLM. Using default plan.")
plan = ["summary_agent", "synthesis_agent", "future_scope_agent"]
print("="*40)
print("π€ DEFAULT ORCHESTRATOR PLAN (LLM call failed)")
print("="*40)
print("\nπ Reasoning: Default plan due to LLM call failure\n")
print("π Planned Steps:")
for i, step in enumerate(plan, 1):
print(f" {i}. {step}")
print("="*40)
print("βοΈ EXECUTE PLAN")
print("="*40 + "\n")
# Create update dict that only modifies necessary fields
updates = {
"plan": plan,
"phase": "EXECUTE"
}
# Only add user_query to messages if it's not already there
if not any(msg[0] == "user_query" for msg in state.messages):
updates["messages"] = [("user_query", state.user_query)]
return Command(goto=plan[0], update=updates)
# If response exists but plan is empty, use default plan
try:
# Enforce summary_agent as the first step if not already included
if not hasattr(response, 'plan') or not response.plan:
print("β οΈ Response from LLM did not contain a valid plan. Using default plan.")
response.plan = ["summary_agent", "synthesis_agent", "future_scope_agent"]
elif response.plan[0] != "summary_agent":
print("β οΈ Enforcing summary_agent as first step in the plan")
response.plan.insert(0, "summary_agent")
print("="*40)
print("π€ ORCHESTRATOR PLAN")
print("="*40)
print(f"\nπ Reasoning:\n{getattr(response, 'reasoning', 'No reasoning provided')}\n")
print("π Planned Steps:")
for i, step in enumerate(response.plan, 1):
print(f" {i}. {step}")
print("="*40)
print("βοΈ EXECUTE PLAN")
print("="*40 + "\n")
# Create update dict that only modifies necessary fields
updates = {
"plan": response.plan,
"phase": "EXECUTE"
}
# Only add user_query to messages if it's not already there
if not any(msg[0] == "user_query" for msg in state.messages):
updates["messages"] = [("user_query", state.user_query)]
return Command(goto=response.plan[0], update=updates)
except Exception as e:
# Final fallback if response processing fails
print(f"β οΈ Error processing LLM response: {str(e)}. Using default plan.")
plan = ["summary_agent", "synthesis_agent", "future_scope_agent"]
# Create update dict that only modifies necessary fields
updates = {
"plan": plan,
"phase": "EXECUTE"
}
# Only add user_query to messages if it's not already there
if not any(msg[0] == "user_query" for msg in state.messages):
updates["messages"] = [("user_query", state.user_query)]
return Command(goto=plan[0], update=updates)
if len(state.plan) == 0 and state.phase == "EXECUTE":
return Command(
goto="final_answer_tool",
update={"phase": "ANSWER"}
)
if state.phase == "EXECUTE":
next_agent = state.plan[0]
remaining_plan = state.plan[1:]
return Command(
goto=next_agent,
update={"plan": remaining_plan}
)
if state.phase == "ANSWER":
return Command(
goto=END,
update={
"phase": "PLAN",
"messages": [("orchestrator_agent", f"\n{state.final_answer}")]
}
)
return Command(goto=END, update={})
# %% [markdown]
# ### Research Agents
# %%
def summary_agent(state : ResearchAgentState) -> Command:
"""Creates concise, structured summaries of research papers."""
if not state.summary:
# Initialize empty summaries
print("The summary agent is processing the papers... π")
research_findings = []
for filename, content in state.research_papers:
# Create a prompt for each paper
system_prompt = f"""You are a research summarization expert. Please read the provided research paper content and create a clear, concise, and structured summary.
Focus on extracting key findings, methodology, and conclusions.
"""
user_prompt = f"""
Paper: {filename}
Content:
{content[:5000]} # Use first 5000 chars to avoid context limits
Please provide a structured summary with key findings, methodology, and conclusions.
"""
response = call_llm(system_prompt, user_prompt, PaperSummary)
# Check if we got a valid response
if response is None:
print(f"β οΈ Failed to summarize paper {filename}. Creating default summary.")
# Create a default summary
finding = {
"title": filename,
"key_findings": ["Error: Could not summarize this paper due to API issues."],
"methodology": "Not available due to API error",
"conclusion": "Not available due to API error",
"source": filename
}
research_findings.append(finding)
else:
try:
# Extract the key findings from the response
finding = {
"title": filename,
"key_findings": response.key_findings if hasattr(response, 'key_findings') else ["No key findings extracted"],
"methodology": response.methodology if hasattr(response, 'methodology') else "Not provided",
"conclusion": response.conclusion if hasattr(response, 'conclusion') else "Not provided",
"source": filename
}
research_findings.append(finding)
except Exception as e:
print(f"β οΈ Error processing summary for {filename}: {str(e)}")
finding = {
"title": filename,
"key_findings": ["Error processing paper summary."],
"methodology": "Error in processing",
"conclusion": "Error in processing",
"source": filename
}
research_findings.append(finding)
print("Paper summaries complete.")
# Add the summaries to the message history
formatted_summaries = []
for paper in research_findings:
findings_text = "\n".join([f"- {finding}" for finding in paper['key_findings']])
formatted_summary = f"""
## {paper['title']}
### Key Findings:
{findings_text}
### Methodology:
{paper['methodology']}
### Conclusion:
{paper['conclusion']}
"""
formatted_summaries.append(formatted_summary)
combined_summary = "\n\n".join(formatted_summaries)
agent_contribution = ("summary_agent", combined_summary)
# Return updates for both summary and messages
return Command(
goto="orchestrator_agent",
update={
"summary": research_findings,
"messages": [agent_contribution]
}
)
else:
# Summaries already exist, just proceed
return Command(goto="orchestrator_agent", update=state)
def synthesis_agent(state : ResearchAgentState) -> Command:
"""Synthesizes the summaries into a cohesive narrative."""
agent_description = state.available_agents.get("synthesis_agent", {})
system_prompt = agent_description.get("system_prompt")
previous_messages = serialize_messages(state.messages)
user_prompt = f"Please synthesize the following research summaries:\n{previous_messages}"
print("The synthesis agent is creating a literature review...")
response = call_llm(system_prompt, user_prompt)
# Handle None response
if response is None:
response_text = "Error: Could not generate synthesis due to API issues."
print("β οΈ Synthesis agent failed - using default response")
else:
response_text = response.content if hasattr(response, 'content') else str(response)
print("Synthesis complete.")
# Only update messages, don't update synthesis_of_findings
return Command(
goto="orchestrator_agent",
update={
"messages": [("synthesis_agent", response_text)]
}
)
def future_scope_agent(state : ResearchAgentState) -> Command:
"""Identifies research gaps and suggests future work."""
agent_description = state.available_agents.get("future_scope_agent", {})
system_prompt = agent_description.get("system_prompt")
previous_messages = serialize_messages(state.messages)
user_prompt = f"Based on the following literature analysis, please identify gaps and suggest future research directions:\n{previous_messages}"
print("The future scope agent is identifying research gaps...")
response = call_llm(system_prompt, user_prompt, FutureScope)
# Handle None response
if response is None:
print("β οΈ Future scope agent failed - using default response")
report_text = "### Identified Research Gaps\n- Error: Could not identify gaps due to API issues.\n\n### Suggested Future Directions\n- Error: Could not suggest directions due to API issues.\n\n### Concluding Synthesis\nError: Could not generate synthesis due to API issues."
else:
try:
report_text = "### Identified Research Gaps\n"
for gap in response.identified_gaps:
report_text += f"- {gap}\n"
report_text += "\n### Suggested Future Directions\n"
for direction in response.suggested_directions:
report_text += f"- {direction}\n"
report_text += f"\n### Concluding Synthesis\n{response.synthesis}"
except Exception as e:
print(f"β οΈ Error processing future scope response: {str(e)}")
report_text = "### Error\nCould not process future scope analysis due to response format issues."
print("Future scope analysis complete.")
# Only update messages, don't update future_directions_report
return Command(
goto="orchestrator_agent",
update={
"messages": [("future_scope_agent", report_text)]
}
)
def critique_agent(state: ResearchAgentState) -> Command:
"""Provides feedback on the generated analysis."""
agent_description = state.available_agents.get("critique_agent", {})
system_prompt = agent_description.get("system_prompt")
previous_messages = serialize_messages(state.messages)
user_prompt = f"Please critique the following research analysis:\n{previous_messages}"
print("The critique agent is reviewing the analysis... π")
response = call_llm(system_prompt, user_prompt)
# Handle None response
if response is None:
response_text = "Error: Could not generate critique due to API issues."
print("β οΈ Critique agent failed - using default response")
else:
response_text = response.content if hasattr(response, 'content') else str(response)
print("Critique complete.")
# Only update the fields that need updating - avoid updating user_query
return Command(
goto="orchestrator_agent",
update={
"critique": response_text,
"messages": [("critique_agent", response_text)]
}
)
def final_answer_tool(state : ResearchAgentState) -> Command[Literal["orchestrator_agent"]]:
"Final answer tool is invoked to formulate a final answer based on the agent message history"
system_prompt = f"""
You're a helpful research assistant and your role is to provide a concise final answer with all the relevant details to answer the user query, based on the provided agent message history.
Structure your response clearly. Use markdown headings for different sections (e.g., ## Synthesized Findings, ## Future Research Directions).
"""
formatted_history = serialize_messages(state.messages)
user_prompt = f"""
---
**Original Task:**
{state.user_query}
---
**Agent Execution History:**
{formatted_history}
---
Compile the final, comprehensive answer for the user based on the history.
"""
response = call_llm(system_prompt, user_prompt)
# Handle None response
if response is None:
final_answer = "Error: Could not generate final answer due to API issues. Please check the logs and try again."
print("β οΈ Final answer tool failed - using default response")
else:
final_answer = response.content if hasattr(response, 'content') else str(response)
if isinstance(final_answer, str):
final_answer = strip_think_blocks(final_answer)
# Only update the final_answer field, not the entire state
return Command(
goto="orchestrator_agent",
update={"final_answer": final_answer}
)
# %% [markdown]
# ## Graph Definition
# %%
def init_state():
"""Initialize the state with default values."""
return ResearchAgentState(available_agents=available_agents)
graph = StateGraph(ResearchAgentState)
graph.add_node("orchestrator_agent", orchestrator_agent)
graph.add_node("summary_agent", summary_agent)
graph.add_node("synthesis_agent", synthesis_agent)
graph.add_node("future_scope_agent", future_scope_agent)
graph.add_node("critique_agent", critique_agent)
graph.add_node("final_answer_tool", final_answer_tool)
# Define the edges
graph.add_edge(START, "orchestrator_agent")
# Fix the parameter name from 'router' to the correct parameter name
graph.add_conditional_edges(
"orchestrator_agent",
lambda state: (
state.plan[0] if state.phase == "EXECUTE" and state.plan
else "final_answer_tool" if state.phase == "ANSWER"
else END
)
)
graph.add_edge("summary_agent", "orchestrator_agent")
graph.add_edge("synthesis_agent", "orchestrator_agent")
graph.add_edge("future_scope_agent", "orchestrator_agent")
graph.add_edge("critique_agent", "orchestrator_agent")
graph.add_edge("final_answer_tool", "orchestrator_agent")
# Compile the graph
graph = graph.compile()
# %% [markdown]
# ## Gradio functions
# %%
def extract_research_papers(
state_dict,
paper_files,
max_iterations: int
) -> tuple[str, Dict, bool]:
"""Extract text from research papers and update state."""
# Create a new ResearchAgentState or update existing one
if isinstance(state_dict, dict):
state = ResearchAgentState(**state_dict)
else:
state = ResearchAgentState()
# Set max_iterations safely
state.max_iterations = max_iterations
if not paper_files:
return "Please upload at least one research paper to analyze.", state.model_dump(), False
console_output = StringIO()
with contextlib.redirect_stdout(console_output):
papers = []
for file in paper_files:
try:
filename = file.name.split("/")[-1]
print(f"π Processing {filename}...")
if filename.lower().endswith(".pdf"):
# Fix DocumentConverter usage - it likely uses a different method name
try:
converter = DocumentConverter()
# Try different method names that might exist
if hasattr(converter, 'pdf_to_text'):
content = converter.pdf_to_text(file.name)
elif hasattr(converter, 'extract_text'):
content = converter.extract_text(file.name)
else:
# Fallback to PyPDF2 if available
import PyPDF2
content = ""
with open(file.name, 'rb') as pdf_file:
pdf_reader = PyPDF2.PdfReader(pdf_file)
for page_num in range(len(pdf_reader.pages)):
content += pdf_reader.pages[page_num].extract_text()
except ImportError:
print("β οΈ PDF conversion libraries not available. Please install PyPDF2.")
continue
elif filename.lower().endswith(".docx"):
doc = docx.Document(file.name)
content = "\n".join([p.text for p in doc.paragraphs])
elif filename.lower().endswith((".txt", ".md")):
with open(file.name, "r") as f:
content = f.read()
else:
print(f"β οΈ Unsupported file format: {filename}")
continue
papers.append((filename, content))
print(f"β
Successfully extracted {len(content)} characters from {filename}")
except Exception as e:
print(f"β Error processing {file.name}: {str(e)}")
state.research_papers = papers
print(f"π Extracted content from {len(papers)} files.")
return console_output.getvalue(), state.model_dump(), len(papers) > 0
def call_orchestrator(state_dict : Dict, user_query : str):
"Function prototype to call the orchestrator agent"
state = ResearchAgentState.model_validate(state_dict)
state.user_query = user_query
buffer = StringIO()
with contextlib.redirect_stdout(buffer):
config = {} # Use empty config for now
try:
result = graph.invoke(input=state, config=config)
output_text = buffer.getvalue()
result_dict = type_conversion(result, ResearchAgentState)
return output_text, result_dict, True
except Exception as e:
error_msg = f"An error occurred during processing: {str(e)}"
output_text = buffer.getvalue() + "\n" + error_msg
return output_text, state_dict, False
result_dict = type_conversion(result, ResearchAgentState)
output_text = buffer.getvalue()
return output_text, result_dict, True
# %% [markdown]
# ## Gradio Interface
# %%
with gr.Blocks() as research_assistant_server:
gr.Markdown("# π Academic Research Assistant")
with gr.Row():
with gr.Column(scale=1): # Image on the far left
try:
gr.Image(value="research_assistant.png", container=False, show_download_button=False, show_fullscreen_button=False)
except Exception: # Broader exception handling
gr.Markdown("*Research Assistant Image*")
with gr.Column(scale=4): # Markdown starts next to the image
gr.Markdown("## Your AI partner for literature reviews and research discovery.")
gr.Markdown("Upload one or more research papers, ask a question, and let the assistant synthesize findings and identify future research directions.")
state_dict = gr.State(value=ResearchAgentState(available_agents=available_agents).model_dump())
extraction_successful = gr.State(value=False)
api_key_set = gr.State(value=API_KEY is not None)
with gr.Tabs():
with gr.TabItem("π API Key Setup"):
gr.Markdown("### Set up your Nebius API Key")
gr.Markdown("A valid API key is required to use this research assistant. You can either provide it here or set it as an environment variable.")
with gr.Row():
nebius_key_input = gr.Textbox(
label="Nebius API Key",
placeholder="Enter your Nebius API key here...",
type="password",
value=""
)
# Add model discovery section
with gr.Row():
discover_models_button = gr.Button("π Discover Available Models", variant="secondary")
test_model_input = gr.Textbox(
label="Or manually test a model name:",
placeholder="e.g., gpt-3.5-turbo"
)
available_models_display = gr.Textbox(
label="Available Models",
lines=5,
interactive=False
)
with gr.Row():
model_dropdown = gr.Dropdown(
choices=NEBIUS_MODELS,
value=MODEL_NAME or NEBIUS_MODELS[0],
label="Select Nebius Model",
allow_custom_value=True
)
api_key_status = gr.Markdown("β οΈ **No API key detected.** Please enter your Nebius API key." if API_KEY is None else "β
**API key configured.** You're ready to use the assistant.")
save_key_button = gr.Button("Save API Key", variant="primary")
def discover_models(key):
if not key:
return "Please enter an API key first."
global API_KEY, ENDPOINT_URL
API_KEY = key
ENDPOINT_URL = "https://api.studio.nebius.com/v1/"
models = list_nebius_models()
if models:
return "Available models:\n" + "\n".join([f"- {model}" for model in models])
else:
return "Could not fetch models. Please check your API key."
discover_models_button.click(
fn=discover_models,
inputs=[nebius_key_input],
outputs=[available_models_display]
)
def save_api_key(key, model):
success = setup_api_key(key, model)
if success:
return f"β
**API key saved successfully!** Using model: {MODEL_NAME}", True
else:
return "β **Invalid API key.** Please check and try again.", False
save_key_button.click(
fn=save_api_key,
inputs=[nebius_key_input, model_dropdown],
outputs=[api_key_status, api_key_set]
)
with gr.TabItem("π Research Materials"):
gr.Markdown("### π Feed the assistant with the research papers you want to analyze.")
with gr.Row():
research_papers_files = gr.File(
label="Upload Research Paper(s)",
file_count="multiple",
file_types=[".pdf", ".txt", ".docx", ".md"],
height=200
)
with gr.Accordion("Advanced options", open=False):
max_iterations = gr.Number(label="Number of refinement iterations", value=1, precision=0)
extract_button = gr.Button("Process Papers", variant="primary")
extract_console_output = gr.Textbox(label="Logs / Console Output")
# Modify extract_research_papers to check for API key
def extract_with_api_check(state_dict, paper_files, max_iterations, api_key_set):
if not api_key_set:
return "β οΈ Please set up your API key in the 'API Key Setup' tab first.", state_dict, False
return extract_research_papers(state_dict, paper_files, max_iterations)
extract_button.click(
fn=extract_with_api_check,
inputs=[state_dict, research_papers_files, max_iterations, api_key_set],
outputs=[extract_console_output, state_dict, extraction_successful]
)
# Rest of your tabs remain the same, but with API key checks for Q&A
with gr.TabItem("π€ Q&A Chatbot"):
examples = """βΉοΈ **Example Queries**
- Summarize the key findings from these papers.
- After synthesizing these articles, what are the main research gaps?
- Propose three future studies based on the provided research.
"""
gr.Markdown(examples)
user_query = gr.Textbox(label="Ask your research question", value="Identify the main gaps and suggest future work.", interactive=True)
button = gr.Button("Ask the Research Assistant π¬π§ ", variant="primary")
# Replace the @gr.render with a proper output textbox
qa_output = gr.Markdown(
label="Research Assistant Response",
value="### π Upload papers and ask a question to get started.",
elem_id="qa_output"
)
output_logs = gr.Textbox(label="Logs/ Console Output", lines=10)
def call_with_api_check(state_dict, user_query, api_key_set):
"""Wrapper to check API key before calling orchestrator."""
if not API_KEY:
error_msg = "β οΈ Please set up your API key in the 'API Key Setup' tab first."
return error_msg, error_msg, state_dict
if not state_dict.get("research_papers"):
error_msg = "### βοΈ No Research Papers Found\n\nππ½ Please upload research papers in the 'π Research Materials' tab first."
return error_msg, error_msg, state_dict
try:
logs, updated_state, success = call_orchestrator(state_dict, user_query)
if success and updated_state.get("final_answer"):
final_answer = updated_state.get("final_answer")
return final_answer, logs, updated_state
else:
error_msg = f"### βοΈ Processing Failed\n\n{logs}\n\nPlease check the logs above for details."
return error_msg, logs, state_dict
except Exception as e:
error_msg = f"### βοΈ An Error Occurred\n\n```\n{str(e)}\n```\n\nPlease check your API key and try again."
return error_msg, f"Error: {str(e)}", state_dict
def reset_output():
"""Reset the output when starting a new query."""
return "### π€ Processing your request...\n\nPlease wait while the research assistant analyzes your papers and generates a response.", "Generating response..."
button.click(
fn=reset_output,
outputs=[qa_output, output_logs]
).then(
fn=call_with_api_check,
inputs=[state_dict, user_query, api_key_set],
outputs=[qa_output, output_logs, state_dict]
)
# with gr.TabItem("π What's under the hood?"):
# gr.Markdown("## Details")
# with gr.Row():
# with gr.Column(scale=1): # Image on the far left
# try:
# gr.Image(value="mas_architecture.png", container=False, label="Architecture")
# except Exception: # Broader exception handling
# gr.Markdown("*Architecture diagram*")
# with gr.Column(scale=2): # Markdown starts next to the image
# gr.Markdown("There is a LangGraph-powered multi-agent system under the hood using an orchestrator approach to plan and route the requests.")
# gr.Markdown("Each agent is specialized in performing academic analysis tasks like summarizing, synthesizing, and identifying future research directions.")
if __name__ == "__main__":
research_assistant_server.launch(mcp_server=True) |