Skip to content

QUICK_DEMO.md

Latest commit

 

History

History
119 lines (84 loc) · 3.17 KB

QUICK_DEMO.md

File metadata and controls

119 lines (84 loc) · 3.17 KB

Quick Demo — first 5 minutes after install

Once ./setup.sh finishes and docker compose ps shows both services Up, walk through this 5-minute tour to verify everything works end-to-end and to see what the assistant can do.

0 — Sanity check (30 s)

In Telegram, message your bot and send:

/help

You should get the command catalog. If nothing comes back, jump to TROUBLESHOOTING.md — "Bot doesn't respond".

1 — Talk to the assistant (1 min)

Pick a project on your machine that OpenCode can see (it lives inside the opencode container's workspace mount). Then in Telegram:

/new
hola, ¿qué proyectos tengo disponibles?

The assistant should respond. It's connected to OpenCode's big-pickle free model by default, and it sees the projects from docker compose exec opencode opencode project list.

2 — Save and recall a fact (1 min)

/memory recordame que prefiero TypeScript estricto

The bot replies Saved fact #N: recordame que prefiero TypeScript estricto.

Now open a brand-new session and ask the assistant:

/new
¿qué lenguaje prefiero?

It should answer with TypeScript. It got that fact from the <known_facts_about_user> block automatically inlined into the system prompt — no tool call needed.

3 — Install a skill from the OpenClaw ecosystem (1 min)

Pick any SKILL.md from https://github.com/alirezarezvani/claude-skills. For example, the git-worktree-manager skill:

/skill_install https://raw.githubusercontent.com/alirezarezvani/claude-skills/main/engineering/git-worktree-manager/SKILL.md

You should see Skill installed: git-worktree-manager. Then:

/listskill

It is now grouped under the engineering category with a ✓ flag. Try its functionality:

/new
list my git worktrees

The assistant should follow the skill's procedure.

4 — See your memory grow (30 s)

/memfiles

You'll see counts of documents (4 — soul, agents, context, session-summary), facts (≥1 from step 2), and skills (≥1 from step 3).

/memory_search Type

Returns the fact you saved.

5 — Confirm the MCP HTTP wiring (1 min)

This is the magic that lets OpenCode update memory mid-session. Run on the host (the machine where docker compose runs):

docker compose exec opencode curl -s -X POST http://bot:4097/mcp \
  -H 'Content-Type: application/json' \
  -d '{"jsonrpc":"2.0","id":1,"method":"tools/list"}' \
  | head -c 200

You should see {"jsonrpc":"2.0","id":1,"result":{"tools":[… — the bot is exposing 9 memory tools and OpenCode can reach them.

What to do next

  • Customize the assistant's identity by editing memory/soul.md on the host, then docker compose restart bot. Personality, tone, hard rules — all live there.
  • Schedule recurring tasks with /task. Examples: a daily summary, a weekly memory backup, a reminder.
  • Read MCP_INTEGRATION.md if you want to understand or extend the memory tools the assistant has.
  • Read TROUBLESHOOTING.md when something breaks.

If any of the five steps above failed, the troubleshooting guide has the symptom-to-fix lookup table.