Manas Documentation

Creating a Research Flow with Specialized Nodes

Intermediate Difficulty schedule 20 minutes

Contents

Building a Research Assistant with Specialized Nodes

This example demonstrates how to create a multi-node flow that acts as a research assistant, using specialized nodes for different tasks in the research process.

Objective

Build a research assistant flow that:

  1. Researches a given topic (Researcher Node)
  2. Analyzes the findings (Analyst Node)
  3. Generates a summary report (Writer Node)

Prerequisites

1

pip install "manas[all-cpu]"

Implementation

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

import os
from core import Flow, LLM
from manas_ai.nodes import QANode

# Initialize LLM (use your preferred provider)
llm = LLM.from_provider(
    "openai",
    model_name="gpt-4",
    api_key=os.environ.get("OPENAI_API_KEY")
)

# Create specialized nodes with different roles
researcher_node = QANode(
    name="researcher",
    llm=llm,
    system_prompt=(
        "You are an expert researcher. Your job is to thoroughly investigate "
        "the given topic and provide comprehensive, factual information. "
        "Include relevant details, statistics, and context."
    )
)

analyst_node = QANode(
    name="analyst",
    llm=llm,
    system_prompt=(
        "You are a data analyst specializing in finding patterns and insights. "
        "Analyze the research provided to you and identify key trends, implications, "
        "and notable findings. Be objective and thorough."
    )
)

writer_node = QANode(
    name="writer",
    llm=llm,
    system_prompt=(
        "You are a skilled technical writer. Create a well-structured summary report "
        "based on the analysis provided. Use clear headings, concise language, and "
        "highlight the most important points. Format your response using Markdown."
    )
)

# Create a flow and connect the nodes
research_flow = Flow()
research_flow.add_node(researcher_node)
research_flow.add_node(analyst_node)
research_flow.add_node(writer_node)

# Connect nodes in sequence
research_flow.add_edge(researcher_node, analyst_node)
research_flow.add_edge(analyst_node, writer_node)

# Process a query through the flow
research_topic = "The impact of artificial intelligence on healthcare diagnostics"
result = research_flow.process(research_topic)

print(f"Research Report on: {research_topic}\n")
print(result)

Explanation

This example demonstrates how to create a multi-node flow with specialized nodes:

  1. Researcher Node: This node is responsible for gathering information on the topic. It uses a system prompt that focuses on thorough research and comprehensive information gathering.

  2. Analyst Node: This node receives the research from the first node and analyzes it to identify key insights, patterns, and implications.

  3. Writer Node: This node takes the analysis and formats it into a well-structured final report with proper formatting and organization.

The flow connects these nodes in sequence, creating a pipeline where the output of each node becomes the input for the next.

Visualization of the Flow

1
2
3
4

┌─────────────┐      ┌─────────────┐      ┌─────────────┐
│ Researcher  │──────► Analyst     │──────► Writer      │
│ Node        │      │ Node        │      │ Node        │
└─────────────┘      └─────────────┘      └─────────────┘

Variations

Adding a Fact-Checker Node

For added reliability, you could add a fact-checking node between the researcher and analyst:

1
2
3
4
5
6
7
8
9
10
11
12
13
14

fact_checker_node = QANode(
    name="fact_checker",
    llm=llm,
    system_prompt=(
        "You are a meticulous fact-checker. Review the research provided "
        "and verify key claims. Flag any potential inaccuracies or "
        "unsubstantiated claims. Provide corrected information where possible."
    )
)

# Add to flow
research_flow.add_node(fact_checker_node)
research_flow.add_edge(researcher_node, fact_checker_node)
research_flow.add_edge(fact_checker_node, analyst_node)

Using a Document Node for External Sources

To incorporate external data sources, you could use a DocumentNode:

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17

from manas_ai.nodes import DocumentNode

document_node = DocumentNode(
    name="document_processor",
    llm=llm,
    system_prompt=(
        "Extract and summarize the key information from the provided documents "
        "that relates to the research topic."
    )
)

# Add relevant documents
document_node.add_document("healthcare_ai_research_paper.pdf")

# Add to flow (parallel to researcher)
research_flow.add_node(document_node)
research_flow.add_edge(document_node, analyst_node)

Complete Example

You can find the complete example in the examples directory of the Manas repository.