General

 

What does the API do?

 

It performs real‑time sentiment analysis on submitted text, returning positive/negative/neutral labels with confidence scores.

​

Is multilingual text supported?

​​

Yes. Language auto‑detection routes content to language‑appropriate models.

​

What security/authentication is used?

 

Access will be via API key on secure REST endpoints,

​

How are errors returned?

 

Structured error responses include codes and messages.

​​

Feeds

​

How do feeds work?

 

The API supports two types of feeds: Platform Feeds and User Feeds.

​

Platform Feeds are ready-to-use, read-only sentiment streams managed by us — no setup required. You can list all available platform feeds or query stats for a specific one (e.g. bitcoin-x). These are available now.

​

User Feeds — where you submit your own text under a custom feedId to build and track your own sentiment stream — are implemented but not yet enabled. Stay tuned!

​

What platform feeds are currently available?

 

Currently one platform feed is live: bitcoin-x, which tracks real-time Bitcoin sentiment from personal opinions posted on X.com. Bots, promotions, and noise are filtered out automatically. More platform feeds are planned.

To explore it:​

• GET /platform-feeds/bitcoin-x — feed metadata and description

• GET /platform-feeds/bitcoin-x/stats — combined sentiment and emotion stats

• GET /platform-feeds/bitcoin-x/stats/sentiment — sentiment-only stats

• GET /platform-feeds/bitcoin-x/stats/emotions — per-emotion stats

​

How does the Bitcoin-X Feed filter posts for sentiment analysis?

 

The Bitcoin-X feed uses a multi-stage pipeline to ensure high-signal, genuine personal sentiment about Bitcoin while aggressively filtering noise, spam, and low-quality content.

​

Here’s how it works:

​

1. Targeted Search We start with a precise X advanced search query focused on “Bitcoin + personal voice” posts, designed to surface opinionated content while minimizing repost chains and obvious noise.

2. Text Normalization Every post is cleaned (URLs, emails, artifacts removed; whitespace and encoding standardized) and malformed/empty entries are dropped early.

3. Fast Rules-Based Filters Immediate rejection of common spam patterns, bot-like content, duplicate-like structures, and low-quality fragments. These run before any LLM call for speed and consistency.

4. Source Credibility Gate Accounts must meet minimum quality signals (follower/activity thresholds) to reduce throwaway or low-signal sources.

5. Aggressive De-duplication Posts are checked against already-ingested content at ingestion time, with a second guard at persistence.

6. Semantic Relevance Check An NLI model verifies that the post is actually about Bitcoin(entailment-based), not just keyword-matched.

7. Intent Gate – Anti-Promo LLM-based classifier rejects commercial/promotional content (referral links, sign-up pushes, overt marketing) in favor of authentic personal commentary.

8. Topic Centrality Gate Ensures Bitcoin is a primary subject, not a passing mention or loose analogy.

9. Personal Voice / Sentiment Gate Final LLM check confirms the author expresses a genuine personal stance or feeling, filtering out neutral reporting, raw price updates, or relayed content.

 

Only posts that pass all gates proceed to sentiment scoring and are added to the feed.

 

Analysis & Processing

​

Can the API detect anomalies or spikes?

​

The API does not currently flag anomalies automatically. However, platform feed stats include bucketed averages and variance for both sentiment and emotions, giving you the raw signal to implement spike detection or threshold alerting on your end. Built-in anomaly flagging is on the roadmap.

​

What metrics are available for monitoring feeds?

​

[Planned] Per‑feed stats will include average sentiment, variance, sample counts, trend direction (improving/worsening/stable), and anomaly indicators over configurable windows.​

​

Can I query by time window?

​

[Planned] You will be able to use parameters like from/to or preset windows (last hour/day/week) to compute metrics and trends for specific periods.​

​

Does the API support batch processing?

 

[Planned] You will be able to submit multiple texts in a single request and receive per‑item sentiment results and errors, which simplifies bulk ingestion.​

​​​

Notifications & Integration

 

Are web-hooks available?

​

[[Planned] You will be able to configure web-hooks for asynchronous processing and notifications, including anomaly events or threshold breaches.​

 

Can I export data?

 

[Planned] You will be able to export feed summaries and aggregates as JSON or CSV for downstream BI/reporting and archival.​