Zero-Entropy Lab
Standard-library research laboratory for deterministic state, evidence, reliability boundaries, agent action integrity, reproducible ledgers, and edge-native execution
Cite this software
Description
Zero-Entropy Lab
Bilingual reading guide / 双语阅读指引(2026-11-13)
中文:本文件属于可靠状态转换、Agent Runtime 容错与执行证据边界的公开仓库说明;查阅当前 src/kernel、实际测试和修订证据,不把概念名称当成运行能力;历史截止点与执行状态不因本次双语整理而改变
English: This root document belongs to fault-tolerant state transitions, agent-runtime contracts, and bounded execution evidence; Check current src/kernel, actual tests, and revision-bound evidence rather than inferring runtime capability from names. Historical cutoffs and execution status are not changed by this bilingual documentation update
Open Research / 开放科研 · Research Template / 科研模板
"Restraint is the ultimate form of digital violence."
A standard-library research laboratory for deterministic state, evidence, reliability boundaries, and edge-native execution.
Status
The Python runtime uses the standard library and keeps generated state reproducible from explicit ledgers.
Zero dependency is an implementation constraint. It is not a claim that inputs, network access, external sources, local state, or generated results are automatically trusted.
Architecture
- Core (
src/kernel/): SQLite-backed memory and deterministic state processing. - Protocol (
src/kernel/protocol/): Local protocol experiments, transition declarations, and lifecycle commands. - Sensory and data inputs (
src/kernel/sensory/,data/inputs/): Explicit source profiles, bounded ingestion, and archive handling. - Reasoning (
src/kernel/cognitive/): Graph-based structural analysis and local heuristics. - Portal (
index.html): Independent presentation surface.
Research and Evidence Surfaces
-
Documentation index — finite navigation for the repository constitution and dated research records. The repository separates implementation, external-source evidence, local state, and research interpretation.
-
aegis-cortex/README.mdis the durable guide to the Aegis reliability/evidence research surface and separates its long-lived policy from time-scoped A1–A6 artifacts. -
aegis-cortex/EVIDENCE_POLICY.mddefines the current reliability-evidence vocabulary and the boundary between external risk, local preventive records, local incidents, execution evidence, and unresolved state. -
data/inputs/ARCHIVE_AND_HARVESTER.mddefines current snapshot, archive, ledger, cache, and ingestion semantics for repository-owned external inputs. -
SECURITY.mddefines the private security-reporting route and supported disclosure boundary. -
RELEASE_POLICY.md,CITATION.cff, andcodemeta.jsondescribe the public software publication and citation surface.
Keep these distinctions explicit:
external risk != local incident
transition declaration != transition execution
hash or HMAC result != source truth or authorization
checker success != semantic validation
repository publication != runtime reproduction
archived publication != later main revision
Long-term evidence and stage status / 长期证据与阶段状态
中文:src/kernel/ 是当前可执行实现;本仓的 Ledger、Aegis、Ballast、周期性研究和外部信源记录各有自己的输入、状态与责任边界;阶段展示应标明来源文件、Git revision、运行者及实际运行时点、覆盖窗口、原始/聚合口径和成功/失败/BLOCKED/NOT_EXECUTED 状态;requirements.txt 是保留的仓库依赖/工具链契约,不能根据其内容看似为空就擅自删除;状态确定性不等于外部来源可信,也不证明任意任务无失败
English: src/kernel/ is current executable implementation; ledgers, Aegis, Ballast, periodic research and external-source records retain distinct inputs, statuses and owners. Stage views must identify source paths, Git revision, producer and actual run time, coverage, raw/aggregate denominator and PASS/FAILED/BLOCKED/NOT_EXECUTED state. The repository's requirements.txt is retained as a tooling/dependency contract, not a file to delete merely because it lists no third-party packages. Determinism does not authenticate external sources or guarantee all tasks succeed
Verification
python tests/run_tests.py
A passing run is evidence for the tested revision, interpreter, environment, and fixtures. It is not a universal reliability or security proof. Record the exact revision and environment when a result is used as reproducibility evidence.
Optional Local Server
The server is a local research interface. It is not required for the lifecycle workflow.
Generate an ephemeral API key before starting it:
export NEXUS_API_KEY="$(python -c 'import secrets; print(secrets.token_urlsafe(32))')"
python src/kernel/protocol/nexus.py serve
The default bind address is 127.0.0.1. API requests fail closed when NEXUS_API_KEY is missing.
A non-loopback bind requires explicit operator opt-in. Do not expose the native server directly to the public Internet. Use an authenticated TLS reverse proxy and network-level access controls when remote access is intentionally enabled.
Security Boundaries
- External documents and API requests remain untrusted inputs.
- Rate limiting is not authentication.
- SHA-256 fingerprints prove content identity, not authorship or authorization.
- Generated graph state must remain reproducible from validated ledgers.
- Local heuristic labels do not establish semantic truth or production guarantees.
- Security reports follow the private process in SECURITY.md.
External Synchronization
External documents are synchronized by src/kernel/sensory/harvester.py from explicit profiles in data/inputs/source_profiles.json.
Historical inputs remain byte-preserved under data/inputs/archive/legacy-through-2026-07-11-1340.
See data/inputs/ARCHIVE_AND_HARVESTER.md for the archive contract.
Citation and Publication Identity
The DOI above identifies an archived Zero-Entropy Lab software publication. Use an exact Git revision in addition to the DOI when a result depends on a specific implementation state.
A DOI is a stable publication identifier. It does not by itself establish scientific validity, local incident evidence, execution success, semantic equivalence with later main, or independent reproduction.
© Zero-Entropy Lab | Built for the Edge, Built for the Future
Archived checkpoint — 2026-10-14
v2026.10-fourteen-day-evidence-checkpoint is archived as 10.5281/zenodo.23365551. For the software-family concept DOI, earlier versions and exact-checkpoint citation, see Release and Archival Policy.
Contributors
Contact person
Xuanyi Jiang
Related projects
Agentic Research Constellation
Ten independent research-software objects connected through evidence, provenance, reproducibility, scientific communication, and longitudinal observation