Skip to main content

Documentation Index

Fetch the complete documentation index at: https://mintlify.com/vrashmanyu605-eng/Langchain_Interview_Multi_Agents_Flow/llms.txt

Use this file to discover all available pages before exploring further.

The project has two configuration surfaces: llm.py, which controls how the agents connect to the language model, and main.py, which sets the input file paths and the initial workflow state. You need to edit both before running the workflow for the first time.

​
LLM configuration (llm.py)

All nine agents share a single ChatOpenAI instance defined in llm.py. The file uses LangChain’s ChatOpenAI class pointed at a locally running LM Studio server, so no API key or external network access is required.
llm.py
from langchain_openai import ChatOpenAI

llm = ChatOpenAI(
    model="gemma-3-4b-it",
    base_url="http://localhost:1234/v1",
    api_key="lm-studio",
    temperature=0.3,
    streaming=True
)
The api_key field is required by LangChain’s ChatOpenAI but LM Studio does not validate it — any non-empty string works.

​
Parameters

​
model
string
default:"\"gemma-3-4b-it\""
The model identifier to request from the LM Studio server. This must match the model name shown in LM Studio’s Local Server tab exactly. Change this whenever you switch to a different model.
​
base_url
string
default:"\"http://localhost:1234/v1\""
The base URL of the OpenAI-compatible server. LM Studio defaults to port 1234. If you changed the port in LM Studio settings, update this value to match.
​
api_key
string
default:"\"lm-studio\""
A placeholder API key. LM Studio accepts any non-empty string. If you point base_url at a real OpenAI-compatible cloud endpoint instead, replace this with your actual API key.
​
temperature
number
default:"0.3"
Controls output randomness. Lower values (closer to 0) make responses more deterministic and focused; higher values (up to 2.0) increase creativity and variation. 0.3 is suitable for structured hiring outputs.
​
streaming
boolean
default:"true"
Enables token-by-token streaming from the LLM. When True, responses begin printing as soon as generation starts. Set to False if you want to receive the full response before processing it.

​
Input configuration (main.py)

The entry point reads two files from disk before building the initial workflow state. Update the two path variables at the top of main.py to point at your files:
main.py
resume_pdf_path = "path/to/your/resume.pdf"
jd_text_path = "path/to/your/job_description.txt"
​
resume_pdf_path
string
required
Absolute path to the candidate’s resume in PDF format. The extract_pdf_text utility opens this file with PyMuPDF (fitz) and concatenates the text content of every page.
​
jd_text_path
string
required
Absolute path to the job description as a plain-text UTF-8 file. The file is read in full and passed directly to the jd_text state field.

​
Initial state configuration

main.py builds the initial_state dictionary that seeds the HiringState TypedDict before the first supervisor call:
main.py
initial_state = {
    "task": "Process candidate hiring workflow",
    "resume_text": resume_text,
    "jd_text": jd_text,
    "workflow_stage": "start",
    "completed": False,
    "workflow_history": []
}
FieldTypeDescription
taskstrHuman-readable description of the workflow goal, visible to agents
resume_textstrAuto-populated from extract_pdf_text(resume_pdf_path)
jd_textstrAuto-populated by reading jd_text_path
workflow_stagestrSet to "start" to signal the supervisor to begin from the first agent
completedboolSet to False; the supervisor sets this to True when the workflow finishes
workflow_historylist[str]Empty list; agents append entries as the workflow progresses

​
Using a different model or endpoint

You can point the system at any OpenAI-compatible endpoint by changing base_url and model in llm.py. For example, to use the OpenAI API directly:
llm.py
llm = ChatOpenAI(
    model="gpt-4o",
    base_url="https://api.openai.com/v1",
    api_key="sk-...",
    temperature=0.3,
    streaming=True
)
Any provider that implements the OpenAI chat completions API — including Ollama, Together AI, and Groq — works the same way. Only base_url, model, and api_key need to change.

Build docs developers (and LLMs) love

Get started for freeTalk to us