Native answers still need an adapter
JevLLM does not pretend Jev generates a free-form investment report. It calls a contract builder and normalizer, and the caller retains responsibility for abstention.
Inspect the supporting source ↗REPOSITORY ANALYSIS
ai-hedge-fund
An educational AI hedge-fund prototype with an optional Jev adapter for structured judgments in fund decision workflows.
E2 · source reviewedPublic material or source evidence was reviewed. A fixed-version source review requires checking a pinned revision; this is not a reproduction, performance endorsement, or safety guarantee.
- Stars
- 63,699
- Forks
- 11,169
- License
- MIT
- Last commit
- 2026-09-22
What Jev does here
Converts strategy questions to System One requests and normalizes native answers to the project’s result format.
Allows Jev or other model backends to be used in the same research workflow.
The author describes an educational research prototype that currently does not execute trades. It is not an official Jev adapter, a zero-hallucination system, or evidence of profitability. This site has not backtested it.
What task does this project fit?
Study how a hedge-fund research workflow can swap a chat provider for native typed decisions.
Material analysed：
From input to action
- Build strategy-specific state and questions.
- JevLLM sends the native request and validates the response.
- Normalize answers and provider metadata into the existing JSON client contract.
Implementation details that matter
Caching includes the actual decision contract
cache_key hashes the agent, built Jev request and contract version. This avoids identifying an answer only by visible prompt text when question semantics change.
Inspect the supporting source ↗When to choose it
Use it to compare decision backends in a research pipeline, not as evidence that typed output makes a strategy profitable or a trade safe.
How to inspect it yourself
Read JevLLM.cache_key and complete together. Check the README's explicit note that the system does not actually make trades against any trading claim.
Limits of this review
- No backtest, fund run or live trade was performed.
- The provider can retry a transient response; adapter latency is not just model inference.