Aspera

A new language for building transparent, adaptive AI

5 followers

A new language for building transparent, adaptive AI

5 followers

Tired of black-box AI? ASPERA merges symbolic logic with LLMs for agents that explain every decision. Declarative DSL, type-safe, adaptive learning. Used in healthcare, fraud detection, customer support. Open-source Python framework.

Free

Launch tags:Developer Tools•Artificial Intelligence•GitHub

Launch Team / Built With

Wispr Flow: Dictation That Works EverywhereStop typing. Start speaking. 4x faster.

Promoted

Maker

📌

👋 **Hey Product Hunt! I’m Christian, creator of ASPERA** I built **ASPERA** to solve a big AI trade-off: • **LLMs are powerful but opaque** — impossible to know *why* a decision was made • **Symbolic systems are transparent but rigid** — zero adaptability 🎯 **ASPERA bridges both worlds natively.** --- ### 💡 What makes it unique * **Declarative syntax** – Write cognitive rules in a clean DSL, no black boxes * **Native explainability** – Every decision produces a full reasoning trace * **Hybrid reasoning** – Symbolic logic for common cases + LLMs for complex context * **Type-safe with LSP** – Autocomplete, linting, and “Go to Definition” as you code * **Plugin ecosystem** – OpenAI, Gemini, Groq, Weaviate, OpenTelemetry * **Adaptive learning** – Agents improve continuously through feedback loops --- ### 🚀 Already in production for: * **Healthcare triage** – Transparent patient prioritization * **Fraud detection** – Clear explanations for red flags * **Customer support automation** – Smart ticket classification and routing --- 📚 **Academic validation:** Published on Zenodo ([DOI: 10.5281/zenodo.17313992](https://doi.org/10.5281/zenodo.1...)) --- ### ⚡ Quick start (5 min) ```bash pip install aspera ``` Create your first agent: ```python # my_agent.aspera agent FraudDetector { goal: "Identify suspicious transactions" rule HighRisk { if amount > 10000 and country in high_risk_list -> flag_for_review reasoning: "Large amount from risky location" } llm_fallback: "Analyze transaction context" model: "gemini-pro" } ``` Run it: ```bash python -m aspera run my_agent.aspera ``` --- ### 🧠 Why I built ASPERA Working on AI for **healthcare and finance**, I realized **explainability isn’t optional**. Decision-makers need to understand *why* an AI chose a particular action — especially in critical domains. --- ### 🗺️ Roadmap 2025 * Visual Agent Builder (drag-and-drop DSL) * Cloud-hosted runtime (zero-config deployment) * Pre-built agent templates for industries * Community plugin marketplace --- 🙏 Happy to answer questions — what would *you* build with ASPERA? 🔗 **Links** * GitHub: [github.com/rthgit/Aspera](https://github.com/rthgit/Aspera) * PyPI: [pypi.org/project/aspera](https://pypi.org/project/aspera/) * Zenodo Paper: [doi.org/10.5281/zenodo.17313992](https://doi.org/10.5281/zenodo.1...) * Docs: [github.com/rthgit/Aspera/docs](https://github.com/rthgit/Aspera...)

Report

11mo ago

Maker

## 🚀 Future: ASTREA Integration (Coming Soon)

Enhanced Performance with ASTREA API:

- Current: 643 searches/sec (built-in fallback)

- Future: 1000+ searches/sec (ASTREA API mode)

- Advanced: Multimodal embeddings (text, images, audio)

- Enterprise: Production-grade semantic engine

ASTREA integration will be available as an optional premium feature for enhanced performance and advanced capabilities.

---

## 🆕 NEW: Semantic Memory

ASPERA agents now have memory!

- ✅ Learn from millions of past decisions

- ✅ Find similar cases instantly (643 searches/sec!)

- ✅ Make smarter decisions based on what worked before

- ✅ Scale to millions of examples without performance loss

### Real Example: Healthcare Agent

```python

agent = AsperaAgent(memory="semantic")

# Learn from 10,000 past diagnoses

agent.learn_from_cases(medical_database)

# New patient with "fever 38.7°C, dry cough"

similar_cases = agent.find_similar(symptoms)

# → Finds "fever 39°C, dry cough" in 3ms

# → Similarity: 92%

# → Past diagnosis: COVID-19 test

# → Success rate: 95%

# Agent recommends same treatment that worked!

```

### Perfect For

- Healthcare: Diagnosis from similar symptoms

- Finance: Trading signals from market patterns

- Legal: Precedent search in case law

- Support: Resolution from similar tickets

### Performance