mkdocs.yml

site_name: AgentDescent
# Written for search, not for us: "RSI" is jargon nobody types into a search box,
# and the phrases people do type -- self-evolving agents, an implementation of a
# named method -- were in none of these pages' metadata.
site_description: >-
  A parallel, asynchronous framework for self-evolving LLM agents, with runnable
  reproductions of twenty-three published self-evolution algorithms.
# Required for canonical URLs, an absolute-URL sitemap, and the social cards
# below -- without it mkdocs emits neither, and the social plugin warns.
site_url: https://thinkflowlab.github.io/agentdescent/
repo_url: https://github.com/ThinkFlowLab/agentdescent
repo_name: ThinkFlowLab/agentdescent
docs_dir: docs
site_dir: site

plugins:
  # `search` is listed explicitly because declaring any plugin replaces the
  # default set, and dropping it silently removes the docs site's own search.
  - search
  # Renders an og:image per page, so a link shared anywhere shows the page title
  # instead of a bare URL. Needs mkdocs-material[imaging]; the docs workflow
  # installs it and the cairo library it loads at runtime.
  - social

theme:
  name: material
  custom_dir: site_overrides
  logo: assets/mark.svg
  font: false
  palette:
    - media: "(prefers-color-scheme: light)"
      scheme: default
      primary: indigo
      accent: indigo
      toggle:
        icon: material/weather-night
        name: Switch to dark mode
    - media: "(prefers-color-scheme: dark)"
      scheme: slate
      primary: indigo
      accent: indigo
      toggle:
        icon: material/weather-sunny
        name: Switch to light mode
  features:
    - navigation.sections
    - navigation.top
    - navigation.tracking
    - navigation.indexes
    - content.code.copy
    - toc.follow

nav:
  - Home: index.md
  # Keep the demo reachable in Material's mobile drawer when header links hide.
  - Interactive demo: https://thinkflowlab.github.io/agentdescent/#demo
  - Documentation: docs.md
  - Getting started:
    - Install and first run: install.md
    - Start here — the plugin, step by step: plugin-quickstart.md
    - Use it from your agent (DSH / Claude Code / Codex / OpenCode / Grok Bot): plugins.md
    - MCP / CLI Path A: mcp-path-a.md
    - Set it up and test it: testing-guide.md
    - Quickstart — dataset to skill: quickstart-skill.md
    - Quickstart — evolve a directory: quickstart-directory.md
    - Quickstart — evolve an agent's code: quickstart-agent-code.md
  - Core concepts:
    - Concepts: concepts.md
    - Architecture: architecture.md
    - The data model: data-model.md
    - Governance (L0 / L1 / L2): governance.md
    - Capability map: capabilities.md
    - The main path: main-path.md
  - The evolve method: evolution.md
  - Modules — design and usage:
    - Module map: modules.md
    - Experimental modules: experimental.md
    - Strategies — what evolves: strategies.md
    - Evolving a directory: directory-evolution.md
    - Agents & LLMs: agents.md
    - Document backends: backends.md
    - The aggregator (the optimizer): aggregator.md
    - Population engine: population.md
    - The ledger: ledger.md
    - The verifier: verifier.md
    - Sparse audit (cheap verifier vs truth): audit.md
    - Statistical evidence: stats-eval.md
    - Decision log: decisions.md
    - Deployment safety — kill/canary/rollback: safety-controller.md
    - RSI metrics — acceptance and cost: metrics.md
    - Diagnose a run: diagnose.md
    - Session controller — when to stop: controller.md
    - Run manifest — reproducibility: manifest.md
    - Evaluation service — budgeted oracle: evaluation.md
    - Safety case — is this archive seedable?: safety.md
    - JEV — a distilled verifier student: jev.md
    - Verifier–curriculum co-evolution: verifier-curriculum.md
    - The data layer: dataloader.md
    - Rewards: rewards.md
    - Reference orchestrator & domain: orchestrator.md
  - Running at scale:
    - Parallelism (DP / TP): parallelism.md
    - Where rollouts run: execution.md
    - Sandboxes: sandboxes.md
    - Async — no round barrier: async.md
    - Duration-aware scheduling: duration-scheduling.md
    - Budget allocation: budget.md
    - The RSI loop kernel: loop-kernel.md
    - Trusted promotion: trusted-promotion.md
    - Trusted evidence: trust-evidence.md
    - Module audit: module-audit.md
    - Adaptive compute: compute-policy.md
    - Capability foundry: capability-foundry.md
    - Continual experience: continual-experience.md
  - Product surfaces:
    - Release a candidate: release.md
    - Trajectory evidence: traces.md
    - Governed harness change: governed-harness.md
    - Tenant cycle: tenant.md
    - Operations console: ops.md
    - The RSI platform cycle: platform.md
    - Capability loop: capability-loop.md
    - Meta campaign: meta-campaign.md
    - Experience store: experience-cli.md
  - The decision plane (policies):
    - Choosing policies: policies.md
    - Using the slots — a guide: policy-guide.md
    - Task sampling: sampling.md
    - Candidate selection: selection.md
    - HGM scheduling: hgm.md
    - Proposal policies: proposal-policies.md
    - Conflict policies: conflict-policies.md
    - Fusion policies: fusion-policies.md
    - Acceptance policies: acceptance-policies.md
    - Leakage guard: leakage.md
    - Marginal-gain stopping: stopping.md
    - Approval gate — high-risk changes: approval.md
    - Edit predictions: predictions.md
    - Guard pipeline: guards.md
    - Guarded runtime: runtime.md
    - Risk policy: risk-policy.md
    - Promotion policies: promotion-policies.md
    - Staleness policies: staleness.md
    - The aggregator_factory exit: aggregator-factory.md
    - Meta-evolution — evolving the slots: meta-evolution.md
  - API reference: api.md
  - Run everything, and extend it: usage.md
  - Examples and results:
    - Complete example — skill evolution: skill-evolution.md
    - Efficiency experiments: efficiency.md
    - Measured results: results.md
    - Runtime matrix: matrix-overview.md
    - Runtime matrix — report: matrix-report.md
    - Molecule search — porous crystals: porous-molecules.md
    - Cross-run consolidation: consolidation.md
    - Harness composition: harness-composition.md
    - Skill compilation: skill-compilation.md
    - Portfolio governance: portfolio.md
    - Portfolio evolution — evolve the spend rule: portfolio-evolve.md
    - Self-improvement loop: self-improve.md
    - Longitudinal RSI report: rsi.md
    - The meta stack on the CLI: meta-cli.md
    - Self-evolution bench — every rung, end to end: self-evolution-bench.md
    - Verified seeding — gate on trust: verified-seed.md
    - Joint multi-slot evolution: joint-meta.md
    - Solver-challenger self-play: selfplay.md
  - Self-evolution algorithms:
    - Overview — twenty-three ports: self-evolution-examples.md
    - Named algorithms (spec / plugin): algorithms.md
    - Register a named Algorithm card: algorithm-authoring.md
    - Algorithm decision tree: algorithm-decision-tree.md
    - Algorithm tournament: tournament.md
    - Campaign — iterate a spec: campaign.md
    - Porting checklist: porting-checklist.md
    - MethodPolicy porting checklist: porting-methodpolicy.md
    - Port fidelity: port-fidelity.md
    - ACE — context evolution: algo-ace.md
    - Dynamic Cheatsheet — test-time memory: algo-dynamic-cheatsheet.md
    - GEPA — prompt evolution: algo-gepa.md
    - EvoSkill — skill discovery: algo-evoskill.md
    - SkillOpt — skill-document training: algo-skillopt.md
    - gskill — repository skill: algo-gskill.md
    - ADAS — meta agent search: algo-adas.md
    - DGM — Darwin Gödel Machine: algo-dgm.md
    - OpenEvolve — program evolution: algo-openevolve.md
    - ERA — empirical-software search: algo-era.md
    - PromptBreeder — genetic prompts: algo-promptbreeder.md
    - AFlow — workflow search: algo-aflow.md
    - Reflexion — episodic memory: algo-reflexion.md
    - Self-Refine — feedback refinement: algo-self-refine.md
    - TextGrad — textual gradients: algo-textgrad.md
    - Voyager — skill library: algo-voyager.md
    - SkillWeaver — web APIs: algo-skillweaver.md
    - Absolute Zero — zero-data self-play: algo-absolute-zero.md
    - R-Zero — challenger/solver: algo-r-zero.md
    - Agent0 — tool curriculum: algo-agent0.md
    - SICA — self-edit: algo-sica.md
    - Gödel Agent — self-modification: algo-godel-agent.md
    - Genesis — persistent recursive worlds: algo-genesis.md
    - RRSI — regularized harness selection: algo-rrsi.md
  - Design records:
    - Evolving a directory: design-directory-evolution.md
    - Plugin for DSH, Claude Code, Codex and other agents: plugin-design.md
    - Meta-evolution — evolving the slots: design-meta-evolution.md

markdown_extensions:
  - admonition
  - meta
  - attr_list
  - md_in_html
  - tables
  - toc:
      permalink: true
  - pymdownx.highlight:
      anchor_linenums: true
  - pymdownx.inlinehilite
  - pymdownx.snippets
  - pymdownx.details
  - pymdownx.superfences:
      custom_fences:
        - name: mermaid
          class: mermaid
          format: !!python/name:pymdownx.superfences.fence_code_format

extra_css:
  - assets/experience.css
  - assets/quickstart.css
  - assets/docs-theme.css
extra_javascript:
  - assets/experience.js

extra:
  social:
    - icon: fontawesome/brands/github
      link: https://github.com/ThinkFlowLab/agentdescent