Using LLM for Question Answering, Text Summarization, and Sentiment Analysis

#learnai #oxlo #ai

Introduction

We are building a lightweight text analysis agent that can summarize long passages, answer questions from a provided context, and score sentiment. It runs entirely through Oxlo.ai's OpenAI-compatible API, so there is no local GPU requirement. Because Oxlo.ai uses flat per-request pricing, feeding it long support tickets or research notes does not drive up cost the way token-based billing would.

What you'll need

Python 3.10+
An Oxlo.ai API key from https://portal.oxlo.ai
The OpenAI SDK: pip install openai

Step 1: Configure the Oxlo.ai client

First, I import the OpenAI SDK and point it at Oxlo.ai's endpoint. This is the only setup required to start routing requests through Oxlo.ai.

from openai import OpenAI

client = OpenAI(base_url="https://api.oxlo.ai/v1", api_key="YOUR_OXLO_API_KEY")

print("Oxlo.ai client ready:", client.base_url)

Step 2: Define the system prompt

I keep the system prompt in a single constant so the behavior is easy to tweak without touching the rest of the logic. It tells the model how to handle each of the three tasks.

from openai import OpenAI

client = OpenAI(base_url="https://api.oxlo.ai/v1", api_key="YOUR_OXLO_API_KEY")

SYSTEM_PROMPT = """You are a precise text analysis engine. You support three operations:
1. Summarization: Condense the input into one concise paragraph.
2. Question Answering: Answer only from the provided context. If the answer is not in the context, say "Not found in context."
3. Sentiment Analysis: Evaluate the text and return a JSON object with exactly two keys: "label" (Positive, Negative, or Neutral) and "score" (an integer 1-10).
Follow the format requested by the user."""

print("System prompt loaded, length:", len(SYSTEM_PROMPT))

Step 3: Implement the three task functions

Next, I add a helper that wraps the Oxlo.ai chat completion call, then three thin functions for summarization, question answering, and sentiment analysis. For sentiment, I enable JSON mode so the output is machine-readable.

import json
from openai import OpenAI

client = OpenAI(base_url="https://api.oxlo.ai/v1", api_key="YOUR_OXLO_API_KEY")

SYSTEM_PROMPT = """You are a precise text analysis engine. You support three operations:
1. Summarization: Condense the input into one concise paragraph.
2. Question Answering: Answer only from the provided context. If the answer is not in the context, say "Not found in context."
3. Sentiment Analysis: Evaluate the text and return a JSON object with exactly two keys: "label" (Positive, Negative, or Neutral) and "score" (an integer 1-10).
Follow the format requested by the user."""

def call_oxlo(user_message: str, json_mode: bool = False):
    kwargs = {
        "model": "llama-3.3-70b",
        "messages": [
            {"role": "system", "content": SYSTEM_PROMPT},
            {"role": "user", "content": user_message},
        ],
        "temperature": 0.2,
    }
    if json_mode:
        kwargs["response_format"] = {"type": "json_object"}
    return client.chat.completions.create(**kwargs)

def summarize(text: str) -> str:
    msg = f"Summarize the following text:\n\n{text}"
    resp = call_oxlo(msg)
    return resp.choices[0].message.content.strip()

def answer_question(context: str, question: str) -> str:
    msg = f"Context:\n{context}\n\nQuestion: {question}"
    resp = call_oxlo(msg)
    return resp.choices[0].message.content.strip()

def analyze_sentiment(text: str) -> dict:
    msg = (
        "Analyze the sentiment of the following text and return only the JSON object. "
        "Do not include markdown formatting.\n\n" + text
    )
    resp = call_oxlo(msg, json_mode=True)
    raw = resp.choices[0].message.content.strip()
    return json.loads(raw)

if __name__ == "__main__":
    print("All three task functions defined and ready.")

Step 4: Wire the runner

Finally, I add a runner block that pushes a realistic support ticket through all three tasks and prints the results. This is the file I actually run from the terminal.

import json
from openai import OpenAI

client = OpenAI(base_url="https://api.oxlo.ai/v1", api_key="YOUR_OXLO_API_KEY")

SYSTEM_PROMPT = """You are a precise text analysis engine. You support three operations:
1. Summarization: Condense the input into one concise paragraph.
2. Question Answering: Answer only from the provided context. If the answer is not in the context, say "Not found in context."
3. Sentiment Analysis: Evaluate the text and return a JSON object with exactly two keys: "label" (Positive, Negative, or Neutral) and "score" (an integer 1-10).
Follow the format requested by the user."""

def call_oxlo(user_message: str, json_mode: bool = False):
    kwargs = {
        "model": "llama-3.3-70b",
        "messages": [
            {"role": "system", "content": SYSTEM_PROMPT},
            {"role": "user", "content": user_message},
        ],
        "temperature": 0.2,
    }
    if json_mode:
        kwargs["response_format"] = {"type": "json_object"}
    return client.chat.completions.create(**kwargs)

def summarize(text: str) -> str:
    msg = f"Summarize the following text:\n\n{text}"
    resp = call_oxlo(msg)
    return resp.choices[0].message.content.strip()

def answer_question(context: str, question: str) -> str:
    msg = f"Context:\n{context}\n\nQuestion: {question}"
    resp = call_oxlo(msg)
    return resp.choices[0].message.content.strip()

def analyze_sentiment(text: str) -> dict:
    msg = (
        "Analyze the sentiment of the following text and return only the JSON object. "
        "Do not include markdown formatting.\n\n" + text
    )
    resp = call_oxlo(msg, json_mode=True)
    raw = resp.choices[0].message.content.strip()
    return json.loads(raw)

if __name__ == "__main__":
    sample = (
        "I have been trying to deploy my container for three hours. "
        "The documentation is unclear about volume mounts, and the CLI keeps returning a 403 error. "
        "I am frustrated because this worked fine last week. "
        "Please help me understand what changed."
    )

    print("=== SUMMARY ===")
    print(summarize(sample))

    print("\n=== ANSWER ===")
    print(answer_question(sample, "What error code is the user seeing?"))

    print("\n=== SENTIMENT ===")
    sentiment = analyze_sentiment(sample)
    print(json.dumps(sentiment, indent=2))

Run it

Save the full script as text_agent.py, swap in your key, and run it.

export OXLO_API_KEY="sk-oxlo.ai-..."
python text_agent.py

When I run this against Oxlo.ai, I see output similar to the following.

Oxlo.ai client ready: https://api.oxlo.ai/v1
=== SUMMARY ===
A user reports difficulty deploying a container due to unclear documentation and a recurring 403 CLI error, noting that the process worked previously and expressing frustration over the change.

=== ANSWER ===
403

=== SENTIMENT ===
{
  "label": "Negative",
  "score": 3
}

Wrap-up and next steps

This agent is already useful for batch-processing documents, but you can extend it. Two concrete next steps: wire it into a FastAPI endpoint so other services can POST text and receive the three analyses, or swap in qwen-3-32b or kimi-k2.6 from Oxlo.ai's catalog if you need stronger multilingual or vision reasoning later.

DEV Community