alphaoptimizer
Codex hook that compacts large command and tool outputs, using Jev to rank which chunks matter and keeping the full output retrievable.
Overview
AlphaOptimizer, from AlphaTales, sits between large command or tool outputs and Codex. Instead of pushing a huge test log, build error, search result or report into the context window, it keeps the useful parts visible, stores the full output locally for a limited time, and lets Codex request more by artifact ID.
Small, failed, structured, media, unsupported and secret-looking outputs pass through unchanged. Eligible text is split into chunks; obvious cases are handled locally, and a small bounded set of normal, non-sensitive chunks goes to Jev for relevance ranking when ALPHAOPTIMIZER_JEV_API_KEY is set. In AlphaTales' initial tests, payloads shrank by about 60 to 65% on average, ranging from about 45% to 80%, though the README stresses these are context reductions, not guaranteed billing savings.
What the decision model decides
| Type | Question | Answers |
|---|---|---|
| Score | How relevant is this chunk of output to the current task?(inferred) | — |
How it works
- Command or tool returns text
- Decide whether to process it (skip small, structured or secret-looking output)
- Split into chunks and send a bounded non-sensitive set to Jev
- Return a compact result to Codex and store the full output locally
- Codex can fetch more by artifact ID; storage expires
Strengths
- Reported 60-65% average payload reduction
- Secret-looking content never sent
- Full output stays retrievable
Caveats
- Codex-only
- Requires trusting hooks and a Jev key for ranking
Getting started
Add the Jev key to Codex's config.toml and trust the AlphaOptimizer hooks, per the README setup.At a glance
- Type
- Service
- Status
- beta
- Language
- Python
- Runtime
- Codex hooks / local service
- Has code
- yes
- License
- MIT
Use cases and integrations
How curated lists describe it
AlphaOptimizer is an open-source tool from AlphaTales that helps Codex work with large command and tool outputs. Instead of sending a huge log or search result straight into the context window, AlphaOptimizer keeps the useful parts visible, keeps the original output available for a limited time, and uses Jev to help rank what matters when an API key is configured.
logicrw/awesome-jev-projects
Frequently asked questions
What does Jev decide?
Which chunks of a large output are relevant enough to show Codex.
Is sensitive output sent to Jev?
No, secret-looking content passes through unchanged and only non-sensitive chunks are sent.
How much does it save?
Initial tests showed about 60-65% payload reduction on eligible outputs.
Related projects
Local compaction proxy for Claude Code and Codex that compacts more often with local rules and keeps full history searchable.
ProxybetaGoClaude Code (and Codex) plugin that trims long Bash output chunk by chunk with Jev before it reaches the model, without summarising.
PluginbetaTypeScriptAgent skill that has Jev pick the few relevant lines from big logs before your coding agent reads them, with a reversible reduce and ledger.
ProjectbetaPythonLocal context-pruning proxy for Claude Code and Codex where Jev decides which history spans can be safely omitted.
ProxyprototypeTypeScriptArchived Go harness finding Jev tool steering does not reliably cut agent tokens in Claude Code or Codex; context trimming helped.
MCP serverresearchGo- winnow★ 110
Calibrated context sieve for Claude Code: Jev judges each block of large tool output and hides confident-no blocks behind recallable stubs.
Projectbeta