OpenAI Cookbook: Official Recipes · Lesson 2

Function Calling for Knowledge Retrieval

Build an agent with two functions: arXiv article search using embeddings, and PDF summarization. The agent decides which function to call based on the user's query.

25 min read3 questions in quizReady prompt includedIn progress

Practical exercise

What to do after this lesson

Add a third function list_downloaded_papers that returns the list of already downloaded papers from the CSV library. Verify the agent uses it when a user asks 'what papers do we already have'.

Ready-to-use prompt

Template for this lesson

Copy and adapt to your context. Text in angle brackets should be replaced.

tools = [
    {
        "type": "function",
        "function": {
            "name": "get_articles",
            "description": "Search arXiv for papers on a topic.",
            "parameters": {
                "type": "object",
                "properties": {
                    "query": {"type": "string", "description": "Search query"}
                },
                "required": ["query"],
            },
        },
    },
]

messages = [
    {"role": "system", "content": "Use functions to answer research questions."},
    {"role": "user", "content": user_question},
]

response = client.chat.completions.create(
    model="gpt-4o-mini", messages=messages, tools=tools
)
# Loop while model wants to call tools
while response.choices[0].finish_reason == "tool_calls":
    call = response.choices[0].message.tool_calls[0]
    result = dispatch(call)  # your dispatcher
    messages += [response.choices[0].message,
                 {"role": "tool", "tool_call_id": call.id,
                  "name": call.function.name, "content": str(result)}]
    response = client.chat.completions.create(
        model="gpt-4o-mini", messages=messages, tools=tools
    )

Defining Functions

from openai import OpenAI from scipy import spatial import pandas as pd GPT_MODEL = "gpt-4o-mini" EMBEDDING_MODEL = "text-embedding-ada-002" client = OpenAI() tools = [ { "type": "function", "function": { "name": "get_articles", "description": ( "Use this function to get academic papers from arXiv " "to answer user questions." ), "parameters": { "type": "object", "properties": { "query": { "type": "string", "description": "User query in plain text", } }, "required": ["query"], }, }, }, { "type": "function", "function": { "name": "read_article_and_summarize", "description": "Summarize a paper already in the library.", "parameters": { "type": "object", "properties": { "query": {"type": "string"} }, "required": ["query"], }, }, }, ]

Ranking by Cosine Similarity

def strings_ranked_by_relatedness( query: str, df: pd.DataFrame, top_n: int = 5, ) -> list[str]: query_embedding = client.embeddings.create( input=query, model=EMBEDDING_MODEL ).data[0].embedding strings_and_scores = [ ( row["filepath"], 1 - spatial.distance.cosine(query_embedding, row["embedding"]), ) for _, row in df.iterrows() ] strings_and_scores.sort(key=lambda x: x[1], reverse=True) return [s for s, _ in strings_and_scores[:top_n]]

Agent Loop

def run_conversation(user_query: str): messages = [ {"role": "system", "content": "Answer using the provided functions."}, {"role": "user", "content": user_query}, ] response = client.chat.completions.create( model=GPT_MODEL, messages=messages, tools=tools ) while response.choices[0].finish_reason == "tool_calls": tool_call = response.choices[0].message.tool_calls[0] name = tool_call.function.name args = json.loads(tool_call.function.arguments) result = get_articles(args["query"]) if name == "get_articles" else read_article_and_summarize(args["query"]) messages.append(response.choices[0].message) messages.append({"role": "tool", "tool_call_id": tool_call.id, "name": name, "content": str(result)}) response = client.chat.completions.create( model=GPT_MODEL, messages=messages, tools=tools ) return response.choices[0].message.content

Сообщить об ошибке

Function Calling for Knowledge Retrieval

Квиз — 3 вопроса

Agent Architecture

Defining Functions

Ranking by Cosine Similarity

Agent Loop