gsp
Valence: macro news sentiment engine scoring global headlines' market impact with a contrastive System-One model on serverless GPU.
Overview
Valence (formerly GSP) is an open-source quantitative macroeconomic sentiment and directional signal engine. It reads global financial and geopolitical news across four equity hubs (US, India, UK and Japan), scores each event's directional market impact with a Contrastive-LM System-One model (CLM-8B) on serverless GPU infrastructure, triggers automated paper-trading signals, and updates a regional macroeconomic knowledge base.
The README argues against generative LLMs for this job, citing format drift, high latency per headline and clustered, arbitrary confidence values, and positions its decision-style model as producing direct probability scores without token decoding. It is a Jev-style approach rather than a Jev integration. It includes CI, deploy and health-check workflows and a Streamlit app. Like other trading demos, its signals are paper trades, not an audited strategy.
What the decision model decides
| Type | Question | Answers |
|---|---|---|
| Choice | What is this headline's directional impact on the market (bullish, bearish, neutral)?(inferred) | — |
| Score | How strong is the expected market impact?(inferred) | — |
How it works
- Ingest global news events
- Score each with the CLM System-One model
- Aggregate into regional signals
- Trigger paper-trading signals and update the knowledge base
Strengths
- No generative decoding for scoring
- Continuous ingestion across regions
- Automated CI and health checks
Caveats
- Uses its own model, not Jev
- Paper trading only; not validated as a strategy
- Promotional README
Getting started
See the Streamlit app at valence.streamlit.app and the repository setup instructions.At a glance
- Type
- Project
- Status
- prototype
- Language
- Python
- Runtime
- serverless GPU + Streamlit
- Has code
- yes
- Has tests
- yes
- Stars
- 0
Use cases and integrations
How curated lists describe it
An AI-driven macro-sentiment analysis terminal that reads thousands of breaking global financial news events in real-time, scores their market impact using a custom Contrastive-LM System-One model on a serverless GPU!
wh000wh000/awesome-jev-live
Frequently asked questions
Does Valence use Jev?
No, it uses its own Contrastive-LM System-One model.
Does it trade real money?
It triggers paper-trading signals.
Which markets?
US, India, UK and Japan.
Related projects
Mangaba AI's local, calibrated typed-decision model for Portuguese text, trained to optimize the Brier score; proprietary since v5.
ProjectproductionPythonCrypto terminal that condenses up to 1,000 posts plus market data into a Jev buy, sell, hold or take-profit card; no trading.
ProjectprototypePythonResearch wiring LAYA/Jev typed decisions into a quant stack with Claude fallback: good news reading, no market-beating edge.
ProjectresearchPythonStandalone Jev-class emotion model: one pass over a message sequence returns calibrated PAD scores, an emotion family and yes/no judgments.
ProjectresearchAudio-native Jev-style model: a frozen Whisper encoder encodes speech once and typed questions return emotion and prosody distributions.
ServiceresearchPythonStudy of whether Jev's reading of financial news predicts stock returns: strong same-day signal, no tradeable alpha.
BenchmarkresearchGo