{"id":50217880,"url":"https://github.com/soturine/soturail","last_synced_at":"2026-05-26T10:01:53.902Z","repository":{"id":359255224,"uuid":"1244990248","full_name":"Soturine/soturail","owner":"Soturine","description":"Local-first context rails for AI coding agents: reversible terminal compression, progressive repo reading, SDD workflows, hooks, benchmarks, memory and cache-friendly payloads.","archived":false,"fork":false,"pushed_at":"2026-05-21T05:42:13.000Z","size":297,"stargazers_count":1,"open_issues_count":0,"forks_count":0,"subscribers_count":0,"default_branch":"main","last_synced_at":"2026-05-21T09:43:47.150Z","etag":null,"topics":["agent-hooks","ai-agents","cli","coding-agents","context-engineering","developer-tools","llm-tools","local-first","prompt-caching","rust","soturail","spec-driven-development","terminal-compression","token-optimization","typescript"],"latest_commit_sha":null,"homepage":"https://github.com/Soturine/soturail#readme","language":"TypeScript","has_issues":true,"has_wiki":null,"has_pages":null,"mirror_url":null,"source_name":null,"license":"mit","status":null,"scm":"git","pull_requests_enabled":true,"icon_url":"https://github.com/Soturine.png","metadata":{"files":{"readme":"README.md","changelog":"CHANGELOG.md","contributing":"CONTRIBUTING.md","funding":null,"license":"LICENSE","code_of_conduct":"CODE_OF_CONDUCT.md","threat_model":null,"audit":null,"citation":null,"codeowners":null,"security":"SECURITY.md","support":null,"governance":null,"roadmap":"ROADMAP.md","authors":null,"dei":null,"publiccode":null,"codemeta":null,"zenodo":null,"notice":null,"maintainers":null,"copyright":null,"agents":"AGENTS.md","dco":null,"cla":null}},"created_at":"2026-05-20T19:56:33.000Z","updated_at":"2026-05-21T05:42:17.000Z","dependencies_parsed_at":null,"dependency_job_id":null,"html_url":"https://github.com/Soturine/soturail","commit_stats":null,"previous_names":["soturine/soturail"],"tags_count":3,"template":false,"template_full_name":null,"purl":"pkg:github/Soturine/soturail","repository_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/Soturine%2Fsoturail","tags_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/Soturine%2Fsoturail/tags","releases_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/Soturine%2Fsoturail/releases","manifests_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/Soturine%2Fsoturail/manifests","owner_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/owners/Soturine","download_url":"https://codeload.github.com/Soturine/soturail/tar.gz/refs/heads/main","sbom_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/Soturine%2Fsoturail/sbom","scorecard":null,"host":{"name":"GitHub","url":"https://github.com","kind":"github","repositories_count":286080680,"owners_count":33514912,"icon_url":"https://github.com/github.png","version":null,"created_at":"2022-05-30T11:31:42.601Z","updated_at":"2026-05-26T03:12:49.672Z","status":"ssl_error","status_checked_at":"2026-05-26T03:12:47.976Z","response_time":63,"last_error":"SSL_read: unexpected eof while reading","robots_txt_status":"success","robots_txt_updated_at":"2025-07-24T06:49:26.215Z","robots_txt_url":"https://github.com/robots.txt","online":false,"can_crawl_api":true,"host_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub","repositories_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories","repository_names_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repository_names","owners_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/owners"}},"keywords":["agent-hooks","ai-agents","cli","coding-agents","context-engineering","developer-tools","llm-tools","local-first","prompt-caching","rust","soturail","spec-driven-development","terminal-compression","token-optimization","typescript"],"created_at":"2026-05-26T10:01:52.900Z","updated_at":"2026-05-26T10:01:53.887Z","avatar_url":"https://github.com/Soturine.png","language":"TypeScript","funding_links":[],"categories":[],"sub_categories":[],"readme":"\u003cp align=\"center\"\u003e\n  \u003cimg src=\"docs/assets/soturail-fox.svg\" alt=\"SotuRail fox logo\" width=\"180\" /\u003e\n\u003c/p\u003e\n\n# SotuRail\n\n[![Node.js \u003e=20](https://img.shields.io/badge/node-%3E%3D20-3c873a)](package.json)\n[![TypeScript](https://img.shields.io/badge/TypeScript-strict-3178c6)](tsconfig.json)\n[![License: Reserved Rights](https://img.shields.io/badge/license-reserved--rights-red)](LICENSE)\n[![CI](https://github.com/Soturine/soturail/actions/workflows/ci.yml/badge.svg)](https://github.com/Soturine/soturail/actions/workflows/ci.yml)\n[![npm version](https://img.shields.io/npm/v/soturail.svg)](https://www.npmjs.com/package/soturail)\n[![npm downloads](https://img.shields.io/npm/dm/soturail.svg)](https://www.npmjs.com/package/soturail)\n[![local-first](https://img.shields.io/badge/local--first-yes-f97316)](docs/security-model.md)\n[![context-engineering](https://img.shields.io/badge/context--engineering-SotuRail-7c3aed)](docs/prompt-caching.md)\n\nSotuRail — Local-first context rails for AI coding agents.  \nSotuRail — trilhos locais de contexto para agentes de IA.\n\n## 1. What Is SotuRail?\n\nSotuRail is a local-first Context OS for AI coding agents such as Claude Code, Codex CLI, Gemini CLI, Cursor, Antigravity-style hosts and similar tools.\n\nIt wraps a repository and terminal session with reversible evidence rails: heuristic repo maps, progressive file reading, safe command execution, raw log recovery, reducers, prompt-cache-friendly blocks, Spec-Driven Development artifacts, local memory, hooks, benchmarks, rules extraction, context packs, agent exports and workflow records.\n\nSotuRail is not the agent, not a Claude-only harness, not a Mermaid-only workflow tool and not a heavy production gateway. It is the local rail layer that helps agents work with better context, safer logs, smaller prompts, approved memory, structured payloads, diagrams, policies and auditable workflows.\n\n## Project Status\n\nv0.5.1 is a polish milestone on top of the v0.5.0 MVP rails. TypeScript mode is stable for local usage. Native Rust mode remains optional and focused on hot paths. Skill Rail, MCP, context packs, agent exports, Workflow Rail, Memory Rail, Context Intelligence, Policy Rail and evidence seeds are local-first and benchmarkable. External comparisons are optional and user-provided.\n\nThe next quality direction is v0.5.2: memory recall quality, context selection quality, role-pack quality, structured payload quality, evidence-pack completeness and diagram validation fixtures. See [ROADMAP.md](ROADMAP.md) and [docs/future-rails-index.md](docs/future-rails-index.md).\n\n## v0.5.x MVP Rails\n\n```bash\nsoturail memory remember \"Decision: keep MCP read-oriented by default\" --tag architecture\nsoturail memory recall \"MCP safety\" --limit 5\nsoturail context select --query \"prepare npm release\"\nsoturail context budget --target claude --explain\nsoturail context pack --role reviewer\nsoturail harness contract init\nsoturail policy doctor\nsoturail fs snapshot\nsoturail mcp exposure\nsoturail run workspace new \"Try v0.5.x rails\"\nsoturail native doctor\nsoturail validate json package.json --strict\nsoturail format compare README.md\n```\n\nThese rails write local JSON, JSONL and Markdown under `.soturail/`. They do not create cloud resources, background agents, global config writes or arbitrary MCP shell execution.\n\n## Why SotuRail Exists\n\nAI coding agents often receive too much unstable context: full files, noisy test logs, repeated terminal output and long conversational summaries. SotuRail is designed to unify those workflows into one independent local-first tool without sending telemetry or inventing provider metrics.\n\nThe long-term product direction is simple:\n\n```txt\nSotuRail does not replace Claude, Codex, Gemini, Cursor or other agents.\nSotuRail prepares the local context, memory, logs, diagrams, policies, payload formats and reports those agents need.\n```\n\n## Where It Fits\n\nUseful mental model:\n\n```txt\nHermes-like systems: agent brain and execution loop.\nDeep Agents-style systems: batteries-included harness with sub-agents, tools, filesystem, memory and approvals.\nClaude Code Harness-style systems: disciplined setup/plan/work/review/release loops with guardrails and evidence.\nMDDD-style systems: Mermaid/.spec.md visual contracts before implementation.\nPlano-like systems: gateway, router and production data plane.\nSotuRail: local Context OS for context, memory, reducers, policy, logs, workflows, diagrams, payload formats and reports.\n```\n\nA newer mental model from the v0.5 planning cycle:\n\n```txt\nDense-agent setup: every task gets every instruction, file and rule.\nSotuRail setup: route the task to the right local context expert, memory, role pack, rule set, payload format, diagram and workflow evidence.\n```\n\nSotuRail absorbs patterns from the context-engineering ecosystem without vendoring or copying adjacent projects. It stays small, local-first, npm-friendly and safe-by-default. Research notes and product ideas live in [docs/ecosystem-influences.md](docs/ecosystem-influences.md), [docs/comparisons.md](docs/comparisons.md), [docs/deep-agents-patterns.md](docs/deep-agents-patterns.md), [docs/future-rails-index.md](docs/future-rails-index.md), [docs/harness-rail.md](docs/harness-rail.md), [docs/policy-rail.md](docs/policy-rail.md), [docs/diagram-rail.md](docs/diagram-rail.md), [docs/structured-payload-rail.md](docs/structured-payload-rail.md) and [docs/agent-docs-hygiene.md](docs/agent-docs-hygiene.md).\n\n## Built With SotuRail\n\nSotuRail dogfoods itself for release-oriented development. `soturail self all` runs repository checks, indexing, build, tests, benchmarks and a local Markdown report through SotuRail's own rails.\n\n```bash\nsoturail self doctor\nsoturail self index\nsoturail self build\nsoturail self test\nsoturail self bench\nsoturail self report\nsoturail self all\n```\n\nReports are written to `.soturail/reports/self-dogfood.md` with stable project context first and dynamic raw IDs, command status and benchmark data later.\n\n## Key Features\n\n- Heuristic Repo Map with cross-platform ignore handling.\n- Progressive reader for large files with reversible collapsed ranges.\n- Safe tee-stream runner with raw log preservation.\n- Git, test, npm, TypeScript, Docker, ESLint, Java/Maven/Gradle, build, JSON and generic terminal reducers.\n- Cross-call and block-level dedupe for repeated command output.\n- Optional Rust native reducer and runner hot paths with TypeScript fallback.\n- Reproducible local benchmark suite.\n- Agent response compression modes.\n- Knowledge-to-Rules ingestion and validators.\n- Prompt-only and hook-style agent integrations.\n- Agent exports for Claude, Codex, Gemini, Cursor, Antigravity and generic hosts.\n- MCP-compatible local stdio server and config helpers.\n- Workflow Rail for local task records and optional worktree planning.\n- Memory Rail with explicit local records, recall, capture and consolidation.\n- Context Intelligence with selection, pruning, budget reports, offload/restore and role packs.\n- Harness, Policy, Filesystem Evidence and Run Workspace seeds.\n- MCP exposure reports and skill routing seeds.\n- Agent docs hygiene checks for short root instruction files.\n- SDD specs, approved memory and cache block normalization.\n- Honest local metrics.\n\n## Planned Next Features\n\nThe next roadmap stage strengthens the rails with evaluation and deeper workflow coverage:\n\n- Memory recall quality fixtures and approved-memory export checks.\n- Context selection, budget, offload and role-pack quality fixtures.\n- Structured Payload Rail validation for Markdown, JSON, tagged blocks, compact tables and Mermaid context.\n- Diagram Rail command seeds and validation fixtures.\n- Harness/evidence quality checks that ensure important paths, commands, errors and security notes survive pruning.\n- Agent Capability Matrix: safer guidance for Claude, Codex, Gemini, Cursor, Antigravity, Deep Agents-style hosts, generic hosts and future hosts.\n- Trace and Report Rails: local records of commands, raw IDs, context packs, memory recall, diagrams, policy decisions and workflow state.\n- Evaluation Suite: token savings plus quality checks, not token savings alone.\n- Future local UI/report mode: HTML reports first, Mermaid rendering and MCP Apps/AG-UI-style event surfaces later.\n\n## Future Rails Documentation\n\nFuture planning is split across focused docs so the roadmap does not become the only source of truth:\n\n- [Future Rails Index](docs/future-rails-index.md)\n- [Harness Rail](docs/harness-rail.md)\n- [Policy Rail](docs/policy-rail.md)\n- [Diagram Rail](docs/diagram-rail.md)\n- [Structured Payload Rail](docs/structured-payload-rail.md)\n- [Agent Docs Hygiene](docs/agent-docs-hygiene.md)\n- [Roadmap Addendum](docs/roadmap-harness-diagram-payload-addendum.md)\n\n## Installation\n\nUse directly with npx:\n\n```bash\nnpx soturail --help\nnpx soturail@latest --help\n```\n\nInstall globally:\n\n```bash\nnpm install -g soturail\nsoturail --help\nsoturail --version\n```\n\nWhen v0.5.1 is published, install that exact version with:\n\n```bash\nnpm install -g soturail@0.5.1\n```\n\nFor local development from source:\n\n```bash\nnpm install\nnpm run build\nnpm link\nsoturail --help\n```\n\nRust is optional. TypeScript builds do not require Rust.\n\n```bash\nnpm run build:native   # optional, requires cargo\nnpm run build:all      # TypeScript + native, requires cargo\n```\n\nnpm package: https://www.npmjs.com/package/soturail\n\n## Quick Start\n\n```bash\nsoturail init\nsoturail index\nsoturail read README.md --query \"quick start\"\nsoturail context pack --target all\nsoturail agents doctor\nsoturail agents export --agent all\nsoturail mcp smoke\nsoturail workflow new \"Try SotuRail\"\nsoturail skills init demo-skill\nsoturail skills validate\nsoturail run npm test\nsoturail expand \u003craw_id\u003e\nsoturail stats\n```\n\nFor an installed clean-folder walkthrough, see [docs/first-real-workflow.md](docs/first-real-workflow.md).\n\n## First Clean Project Flow\n\n```bash\nsoturail init\nsoturail memory doctor\nsoturail context budget --explain\nsoturail context pack --role planner\nsoturail harness doctor\nsoturail policy doctor\nsoturail run workspace new \"Try SotuRail\"\nsoturail context pack --target all\nsoturail agents doctor\nsoturail agents export --agent all\nsoturail mcp smoke\nsoturail workflow new \"Try SotuRail\"\nsoturail workflow list\n```\n\n`soturail init` scaffolds agent, MCP, context pack, hook, skill and workflow examples without overwriting user-edited files.\n\n## Commands\n\n```bash\nsoturail init\nsoturail index\nsoturail read \u003cfile\u003e --query \"goal\"\nsoturail run --engine auto \u003ccommand...\u003e\nsoturail run --similar-dedupe conservative npm test\nsoturail run --engine native \u003ccommand...\u003e\nsoturail expand \u003craw_id\u003e\nsoturail expand \u003craw_id\u003e --allow-raw --yes\nsoturail dedupe stats\nsoturail bench prepare\nsoturail bench run --engine ts\nsoturail hooks list\nsoturail agents list\nsoturail agents doctor\nsoturail agents export --agent all\nsoturail mcp smoke\nsoturail mcp config --agent generic\nsoturail context pack --target all\nsoturail context pack --role planner\nsoturail context select --query \"release checklist\"\nsoturail context budget --explain\nsoturail workflow list\nsoturail workflow show \u003cid\u003e\nsoturail workflow close \u003cid\u003e\nsoturail run workspace show \u003crun-id\u003e\nsoturail format README.md --mode concise\nsoturail format compare README.md\nsoturail validate json package.json --strict\nsoturail ingest README.md --type docs\nsoturail rules check\nsoturail skills init demo-skill\nsoturail skills validate\nsoturail skills export --target claude\nsoturail spec status\nsoturail memory propose \"decision\"\nsoturail doctor cache\n```\n\n## Release Workflow\n\nRelease automation is local-first and conservative:\n\n```bash\nnpm run release:check\nnpm run release:publish -- X.Y.Z\nnpm run release:github -- X.Y.Z\nnpm run release:full -- X.Y.Z --publish-npm --github-release\n```\n\nThe release commands accept a positional version, `--target-version X.Y.Z`, or the backward-compatible `--version X.Y.Z` option. The release script never runs `npm audit fix --force`, never publishes when build/tests/runtime audit fail and never creates a GitHub release before npm publish succeeds. See [docs/release-workflow.md](docs/release-workflow.md).\n\nRelease verification installs the packed `.tgz` into a clean temporary project and executes `node_modules/soturail/dist/cli.js` directly to avoid npm cache or global CLI false positives.\n\n```bash\nnpm run build\nnpm test\nnpm run release:check\nnpm pack --dry-run\n```\n\n## Agent Integrations\n\nSotuRail provides reviewed agent integration profiles for Claude, Codex, Gemini, Cursor, Antigravity and generic agents.\n\n```bash\nsoturail agents list\nsoturail agents doctor\nsoturail agents export --agent all\nsoturail agents install --agent claude --mode mcp --dry-run\nsoturail agents install --agent cursor --mode rules --dry-run\n```\n\nExports live under `.soturail/exports/agents/`. Install commands are dry-run-first, backup-first and project-local by default. SotuRail does not auto-enable arbitrary shell execution or unknown global host configs.\n\n## MCP And Context Packs\n\nSotuRail exposes local context through a minimal MCP-compatible stdio server and cache-friendly context packs.\n\n```bash\nsoturail mcp doctor\nsoturail mcp manifest\nsoturail mcp serve --transport stdio\nsoturail mcp smoke\nsoturail context pack --target claude\nsoturail context pack --target codex\nsoturail context pack --target gemini\nsoturail context pack --target cursor\nsoturail context pack --target antigravity\nsoturail context pack --target generic\nsoturail context pack --target all\nsoturail context explain\n```\n\nMCP does not expose arbitrary shell execution. Raw log expansion redacts probable secrets unless raw output is explicitly requested.\n\nContext packs are written to `.soturail/context/\u003ctarget\u003e-context.md`. JSON-RPC examples live under [examples/mcp](examples/mcp).\n\nRole packs, structured payload validation and offload flows are tracked in [docs/context-packs.md](docs/context-packs.md), [docs/structured-payload-rail.md](docs/structured-payload-rail.md) and [docs/deep-agents-patterns.md](docs/deep-agents-patterns.md).\n\n## Workflow Rail\n\nWorkflow Rail stores local task state under `.soturail/workflows/` and can optionally plan Git worktree isolation.\n\n```bash\nsoturail workflow new \"Implement feature\"\nsoturail workflow list\nsoturail workflow show \u003cid\u003e\nsoturail workflow close \u003cid\u003e\nsoturail workflow start \u003cid\u003e --worktree --dry-run\nsoturail workflow verify \u003cid\u003e\n```\n\nSotuRail does not push, merge or delete worktrees automatically.\n\nFuture role-phase workflows, harness-style phases and diagram-driven workflows are tracked in [docs/workflow-rail.md](docs/workflow-rail.md), [docs/harness-rail.md](docs/harness-rail.md) and [docs/diagram-rail.md](docs/diagram-rail.md).\n\n## Skill Rail\n\nSkill Rail creates, validates, exports and packs safe local agent skills without depending on external skill ecosystems.\n\n```bash\nsoturail skills init demo-skill\nsoturail skills list\nsoturail skills validate\nsoturail skills export --target claude\nsoturail skills pack --format markdown\n```\n\nExports are written under `.soturail/exports/skills/` and should be reviewed before enabling in an agent host. Examples live under [examples/skills](examples/skills).\n\nFuture task/role-aware skill routing is tracked in [docs/skill-rail.md](docs/skill-rail.md).\n\n## Reducers And Dedupe\n\nv0.3.2 added stronger reducers for common developer commands including `npm install`, `npm test`, Vitest, `tsc`, `git diff`, `git status`, Docker logs, ESLint, Vite/Next build output and Java/Maven/Gradle failures. Reducers preserve errors, warnings, file paths, line/column references, stack traces, security warnings and the raw recovery hint.\n\nBlock-level dedupe is conservative by default. It can replace repeated safe blocks with references such as `[deduped block: ...]`, while preserving error blocks and current failure context. Similar-output dedupe is experimental and opt-in:\n\n```bash\nsoturail run --similar-dedupe conservative npm test\nsoturail dedupe stats\n```\n\n## Benchmarks And Honest Metrics\n\nSotuRail includes deterministic local fixtures for terminal compression, agent response compression, JSON/tool payload compression, Knowledge-to-Rules structuring, dedupe and native performance readiness.\n\n```bash\nnpm run build\nsoturail bench prepare\nsoturail bench run --engine ts\nsoturail bench report\n```\n\nThe latest report is written to [benchmarks/reports/latest.md](benchmarks/reports/latest.md). External comparisons are optional and user-provided only.\n\nLocal token counts are deterministic estimates. SotuRail reports raw payload tokens, reduced payload tokens, metadata overhead and net estimated tokens. For tiny outputs, compression may be ineffective once recovery metadata is included; SotuRail says that directly while preserving raw recovery paths.\n\nFuture benchmarks will also cover context quality, role-pack quality, diagram validation, evidence-pack completeness and format quality for Markdown vs tagged vs JSON vs compact payloads.\n\n## Knowledge-to-Rules Engine\n\nSotuRail can ingest Markdown, TXT, JSON and YAML into structured YAML/JSON rules, checklists, citations and simple validators.\n\n```bash\nsoturail ingest docs/requirements.md --type requirements\nsoturail rules list\nsoturail rules check\nsoturail rules export --format yaml\n```\n\nThis makes heavy docs smaller, reusable and auditable without inventing rules that do not appear in the source. See [docs/knowledge-to-rules.md](docs/knowledge-to-rules.md), [docs/rules.md](docs/rules.md) and [docs/policy-rail.md](docs/policy-rail.md).\n\n## Prompt Caching Design\n\nStable blocks are ordered before dynamic data:\n\n1. static_header\n2. governance\n3. config\n4. repo_map\n5. approved_specs\n6. approved_memory\n7. dynamic_footer\n\nSotuRail reports estimated cache stability only. It never claims real provider cache hits unless imported metadata exists.\n\n## Native Performance Path\n\nTypeScript remains the public CLI, orchestration, docs and npm distribution layer. Rust handles optional hot paths where streaming, low overhead and binary execution matter:\n\n- native terminal reducers;\n- native JSON/tool payload reducer;\n- native tee-stream runner with raw log and summary sidecar support.\n\nSotuRail does not claim native speedups unless local benchmark results show them for your machine.\n\n## Security Model\n\n`soturail run` blocks dangerous patterns by default, including `rm -rf`, `sudo`, `format`, `dd if=`, `curl | sh`, `del /s` and `git push`.\n\nRaw logs may contain secrets because they preserve real terminal output. Treat `.soturail/raw/` as local evidence, not public artifact material. `soturail expand \u003craw_id\u003e` redacts probable secrets by default; use `--allow-raw --yes` only when you intentionally need exact raw output.\n\nPolicy queue, auth guidance and MCP exposure reports are tracked in [docs/policy-rail.md](docs/policy-rail.md) and [docs/security-model.md](docs/security-model.md).\n\n## Migration\n\nMoving from v0.4.x to v0.5.x keeps the agent export, MCP and Workflow Rail commands, then adds local `.soturail/` folders for Memory, Context Intelligence, Harness, Policy, Filesystem Evidence and Run Workspace seeds. See [docs/migration-v0.5.md](docs/migration-v0.5.md).\n\n## Windows Notes\n\nWindows users should see [docs/windows.md](docs/windows.md) for CMD vs PowerShell quoting, global install, `npx`, local tarball testing and common paste mistakes such as copying Markdown code-fence labels into CMD.\n\n## Comparison Philosophy\n\nSotuRail is inspired by the broader context-engineering ecosystem, including terminal reducers, agent response compression, spec-driven workflows, local memory, rules extraction, hooks, benchmarks, skill registries, agent memory, harness workflows, Mermaid diagram-driven development, structured prompt payloads and gateway/observability ideas. SotuRail does not vendor or depend on those projects. It aims to unify similar ideas into one local-first workflow while keeping benchmarks honest. See [docs/comparisons.md](docs/comparisons.md), [docs/ecosystem-influences.md](docs/ecosystem-influences.md), [docs/deep-agents-patterns.md](docs/deep-agents-patterns.md) and [docs/future-rails-index.md](docs/future-rails-index.md).\n\n## Roadmap\n\nSee [ROADMAP.md](ROADMAP.md).\n\nNear-term direction:\n\n```txt\nv0.5.0  Memory Rail + Context Intelligence + Role Packs + Harness/Policy seeds + reliability\nv0.5.1  Memory/context polish + Structured Payload Rail + Agent Docs Hygiene + Diagram docs\nv0.5.2  Evaluation suite + context/role-pack/format/evidence/diagram quality benchmarks\nv0.6.0  Real agent runtime integration + host capability matrix + host-aware payload/policy docs\nv0.7.0  Workflow Rail 2.0 + Harness Rail + Diagram Rail + .spec.md visual contracts\nv0.8.0  Knowledge Rail and Project Brain with specs, diagrams and recurring failures\nv0.9.0  Native Engine Real\nv0.10.0 Local reports, traces, Mermaid rendering and dashboard\nv1.0.0  Stable Context OS\n```\n\n## Contributing\n\nSee [CONTRIBUTING.md](CONTRIBUTING.md). Reducers, hooks and rules should include tests, safety notes and benchmark impact when relevant.\n\n## License Status\n\nSotuRail is not currently licensed as open source for the current repository state or future versions.\n\nAll rights are reserved unless a later LICENSE file or written agreement explicitly grants permissions. This repository may be public for portfolio, planning, research, documentation and evaluation visibility, but broad copying, redistribution, sublicensing, commercial use or derivative works are not granted by default.\n\nImportant: older SotuRail versions or copies that were explicitly released under MIT remain governed by the MIT terms that applied to those specific versions at the time of release.\n\nSee [docs/licensing-strategy.md](docs/licensing-strategy.md).\n","project_url":"https://awesome.ecosyste.ms/api/v1/projects/github.com%2Fsoturine%2Fsoturail","html_url":"https://awesome.ecosyste.ms/projects/github.com%2Fsoturine%2Fsoturail","lists_url":"https://awesome.ecosyste.ms/api/v1/projects/github.com%2Fsoturine%2Fsoturail/lists"}