Skip to main content
Classify sentiment in text — customer reviews, social media posts, support tickets — using structured output that returns typed JSON with scores and categories.

​
Single text analysis

from runcrate import Runcrate
import json

client = Runcrate(api_key="rc_live_YOUR_API_KEY")

review = "The laptop display is gorgeous. Battery life could be better — barely lasts 5 hours."

response = client.models.chat_completion(
    model="deepseek-ai/DeepSeek-V3",
    messages=[
        {"role": "system", "content": "Analyze sentiment. Return JSON: {sentiment, score (-1 to 1), topics, summary}"},
        {"role": "user", "content": review},
    ],
    max_tokens=256,
    temperature=0.1,
    response_format={"type": "json_object"},
)

result = json.loads(response.choices[0].message.content)
print(f"Sentiment: {result['sentiment']} ({result['score']})")
print(f"Topics: {', '.join(result['topics'])}")

​
Batch analysis — customer reviews

from runcrate import Runcrate
from concurrent.futures import ThreadPoolExecutor
import json

client = Runcrate(api_key="rc_live_YOUR_API_KEY")

reviews = [
    {"id": 1, "text": "Absolutely love this product. Best purchase this year."},
    {"id": 2, "text": "Arrived broken. Support took 3 weeks to respond."},
    {"id": 3, "text": "It's fine. Does what it says, nothing special."},
    {"id": 4, "text": "Great build quality but the software is buggy."},
]

def analyze(review):
    response = client.models.chat_completion(
        model="deepseek-ai/DeepSeek-V3",
        messages=[
            {"role": "system", "content": "Return JSON: {sentiment, score, topics, summary}"},
            {"role": "user", "content": review["text"]},
        ],
        max_tokens=256, temperature=0.1,
        response_format={"type": "json_object"},
    )
    return {"id": review["id"], **json.loads(response.choices[0].message.content)}

with ThreadPoolExecutor(max_workers=5) as pool:
    results = list(pool.map(analyze, reviews))

avg_score = sum(r["score"] for r in results) / len(results)
print(f"Average score: {avg_score:.2f}")
for r in results:
    print(f"  #{r['id']}: {r['sentiment']} ({r['score']}) — {r['summary']}")

​
Next.js API route

// app/api/analyze-sentiment/route.ts
import { runcrate } from '@runcrate/ai';
import { generateText, Output } from 'ai';
import { z } from 'zod';

const SentimentResult = z.object({
  sentiment: z.enum(['positive', 'negative', 'neutral', 'mixed']),
  score: z.number().min(-1).max(1),
  topics: z.array(z.string()),
  summary: z.string(),
});

export async function POST(req: Request) {
  const { text } = await req.json();

  const { output } = await generateText({
    model: runcrate('deepseek-ai/DeepSeek-V3'),
    output: Output.object({ schema: SentimentResult }),
    prompt: `Analyze the sentiment of this text:\n\n${text}`,
  });

  return Response.json(output);
}

​
Social media monitoring

from runcrate import Runcrate
import json

client = Runcrate(api_key="rc_live_YOUR_API_KEY")

posts = [
    "Just tried @acme's new app — the onboarding is so smooth.",
    "@acme your checkout is broken AGAIN. Third time this month.",
    "The new @acme update fixed everything. Finally works on my phone.",
]

for post in posts:
    response = client.models.chat_completion(
        model="deepseek-ai/DeepSeek-V3",
        messages=[
            {"role": "system", "content": "Return JSON: {sentiment, score, urgency: low|medium|high, action_needed: bool}"},
            {"role": "user", "content": post},
        ],
        max_tokens=128, temperature=0.1,
        response_format={"type": "json_object"},
    )
    r = json.loads(response.choices[0].message.content)
    flag = " [ACTION]" if r.get("action_needed") else ""
    print(f"{r['sentiment']:>8} ({r['score']:+.1f}) | {post[:55]}...{flag}")

​
Tips

  • Temperature 0.1 gives the most consistent classifications.
  • response_format: json_object forces JSON output from the API.
  • Vercel AI SDK Output.object gives typed results with Zod validation.
  • DeepSeek-V3 handles nuanced sentiment — it catches mixed reviews and sarcasm.

​
Next steps