For trading bots, dashboards & research workflows

Turn market news into sentiment your trading system can use.

Analyse news across your watchlist and get structured sentiment scores, confidence and source information through one API. Add news context to your trading bots, market dashboards and research workflows.

See an example Deploy TierPulse

Open source Self-hosted Local FinBERT analysis

News in. Structured sentiment out.

Illustrative example

1. Send your watchlist

AAPL / TSLA / XYZ

2. Analyse available news

AAPL Earnings beat expectations as demand rises

TSLA Deliveries fall short of forecasts

XYZ No news available

3. Use the results in your system

Illustrative sentiment results with news-only mode enabled
Symbol	Sentiment	Score	Confidence
AAPL	Positive	+0.82	91%
TSLA	Negative	−0.64	87%
XYZ	Unavailable	—	0%

Example uses news-only mode. Available results use local FinBERT analysis; missing news is marked unavailable, so your system can handle it explicitly.

Trading bots Dashboards Research

The Intelligence Failover Strategy

Trading bots cannot afford downtime or hallucinations. TierPulse ensures reliability by cascading analysis across three distinct layers.

High-Scale Batching

Processes millions of tokens per second with Rust's memory safety. Automatically batches requests for news providers and LLMs to minimize latency.

Multi-Tier Caching

Hybrid caching strategy using Moka for fast in-memory lookups and Redis for distributed coordination. Scalable performance without bottleneck.

Traffic Control

Built-in "Token Bucket" rate-limiter protects upstream API quotas. Ensures "Five Nines" (99.999%) availability for critical missions.

Why Rust?

Memory Safety: Elimination of data races in high-concurrency sentiment streams.

Zero-Bloat Deployment: Multi-stage distroless builds resulting in a container footprint around 400MB.

Targeted Availability: Engineering towards "Five Nines" (99.999%) uptime for institutional missions.

// TierPulse Engine Orchestration Layer
pub struct AppState {
  engine: InferenceEngine, // Tier 1: Local ONNX
  limiter: RateLimiter, // governor protection
  cache: Cache<String, SentimentResult>,
}

impl AppState {
  pub async fn analyze(&self, req: AnalyzeRequest) -> Result<Response, Error> {
    // 1. Enforce Traffic Control
    self.limiter.check().await?;

    // 2. Sequential Intelligence Failover
    if let Some(cached) = self.cache.get(&req.ticker).await {
      return Ok(cached);
    }

    // Escalation Tier 2 (Providers) -> Tier 3 (LLMs)
    self.fetch_with_failover(req).await
  }
}

Deploy with Compose

The recommended way to deploy TierPulse is using Docker Compose. A ready-to-use docker-compose.yml is included at the repository root for quick orchestration of the engine and local cache layers.

Recommended: Launch Stack

docker compose up -d

Verify Health

curl -s http://localhost:8080/health/ready

View on Docker Hub

Quick Start Environment (.env)

TP_TIINGO_KEY=your_key_here

TP_AUTH_MODE=api_key

TP_AUTH_API_KEYS=tenant:key

TP_PRIMARY_LLM=grok

# Optional: TP_REDIS_URL, TP_GROK_KEY, TP_DEEPSEEK_KEY, etc.

API Request Payload

{
  "symbols": [
    { "ticker": "AAPL", "name": "Apple Inc." },
    { "ticker": "TSLA", "name": "Tesla, Inc." },
    { "ticker": "BTC", "name": "Bitcoin" }
  ],
  "lookback_hours": 24,
  "max_articles_per_symbol": 5
}

POST /api/v1/analyze

Standardized Response

{
  "request_id": "tp_550e8400-e29b-41d4-a716-446655440000",
  "results": [
    {
      "symbol": "AAPL",
      "sentiment_score": 0.82,
      "label": "bullish",
      "confidence": 0.94,
      "source_tier": "tier_1_local_onnx"
    },
    {
      "symbol": "BTC",
      "sentiment_score": -0.45,
      "label": "bearish",
      "confidence": 0.88,
      "source_tier": "tier_3_llm",
      "reasoning": "Recent regulatory tightening in EU..."
    }
  ],
  "execution_time_ms": 450
}

JSON Response (HTTP 200)