Wikikiki

A wiki that is the live memory of a system of actors — human, agent, and machine — observable and editable in real time.

See DESIGN.md for the full architecture document.

Status

Stage 1 — markdown wiki with git backing, full access foundation, two actor types (human via cookie session, agent via API token), attribution, FTS5 search.

Stages 2–4 (live working memory, distillation, hardening) are deferred per §17.

Quickstart

# 1. Build
cargo build --release

# 2. Configure (optional — defaults are fine)
cp wikikiki.example.toml wikikiki.toml

# 3. Bootstrap the first admin (creates the wiki + DB + git repo)
./target/release/wikikiki bootstrap --handle alice --password 'change-me'

# 4. Run the server
./target/release/wikikiki serve

# 5. Open http://localhost:8090 and log in as alice.

Issuing an agent token

./target/release/wikikiki issue-token --handle research-bot --type agent --label "first agent"
# prints the plaintext token once; store it somewhere safe.

Use it from an agent:

curl -H "Authorization: Bearer <token>" http://localhost:8090/api/pages/welcome
curl -H "Authorization: Bearer <token>" -X PUT \
     -H "Content-Type: text/markdown" \
     --data 'hello from an agent' \
     http://localhost:8090/api/pages/notes/from-agent

Layout

src/
├── main.rs        CLI entrypoint (serve, bootstrap, issue-token)
├── lib.rs
├── config/        TOML loader with env overrides
├── db/            sqlx pool, migrations runner, queries
├── git/           git2 wrapper for the content repo
├── web/           axum router, maud templates, auth middleware
└── actor/         REST API surface for agents
migrations/        sqlx SQL migrations (lookups seeded here)
content/           the wiki itself (markdown files; path is configurable)

---

A part of the TAGENT Private Edge Office Stack | An AI Agent Platform Project by AI VISIONS

https://www.linkedin.com/feed/update/urn:li:activity:7447711407745200129/