Merge pull request #7 from lneely/update-docs
Tech edit README: improve clarity, concision, and flow
This commit is contained in:
Levi Neely
GitHub
commit
1b38cbb030
|
|
@ -1,6 +1,12 @@
|
|||
# Events System
|
||||
|
||||
The events system provides a real-time text stream of all AnviLLM activity.
|
||||
The events system provides a real-time text stream of all AnviLLM activity using [github.com/simonfxr/pubsub](https://github.com/simonfxr/pubsub) as the event bus.
|
||||
|
||||
## Architecture
|
||||
|
||||
<p align="center"><img src="diagrams/events.svg?v=2" width="400"></p>
|
||||
|
||||
Multiple clients can read from `agent/events` simultaneously, each consuming the same event stream for different purposes: notifications (desktop, SMS, WhatsApp), logging/auditing, custom integrations, etc.
|
||||
|
||||
## Reading Events
|
||||
|
||||
|
|
|
|||
148
README.md
148
README.md
|
|
@ -1,16 +1,18 @@
|
|||
# AnviLLM
|
||||
|
||||
9P-based LLM orchestrator — Front-ends: Acme, Emacs, TUI, web
|
||||
LLM orchestrator using 9P — scriptable, multi-backend, crash-resilient
|
||||
|
||||
## Architecture
|
||||
|
||||
**Core:** anvilsrv (9P daemon), anvilmcp (MCP server)
|
||||
**Clients:** Assist (Acme), anvillm.el (Emacs), anvillm (TUI), anvilweb (web)
|
||||
**Benefits:** Shared sessions, crash-resilient, scriptable via 9P, service integration, MCP support, **cross-backend agent communication**
|
||||
|
||||
|
||||
**Core:** anvilsrv (9P daemon), anvilmcp (MCP server), anvilwebgw (API proxy)
|
||||
**Clients:** Assist (Acme), anvillm.el (Emacs), anvillm (TUI), anvilweb (web)
|
||||
**Benefits:** Shared sessions, crash recovery, scriptable via 9P, cross-backend agent communication
|
||||
|
||||
## Why 9P?
|
||||
|
||||
AnviLLM uses 9P because orchestration becomes file manipulation (`read`, `write`, `ls`, `stat`). Benefits: simplicity, universal access (shell, C, Go, Python, `cat`, `echo`), and composability (Unix pipes, `grep`, `awk`, `jq`). The only requirement is a 9P client (plan9port's `9p` or compatible).
|
||||
9P turns orchestration into file operations (`read`, `write`, `ls`, `stat`). Control agents with standard tools: `cat`, `echo`, shell scripts, or any language with file I/O. Compose workflows with Unix pipes, `grep`, `awk`, `jq`. Only requirement: a 9P client (plan9port's `9p` or compatible). Built with [9fans.net/go](https://9fans.net/go).
|
||||
|
||||
## Requirements
|
||||
|
||||
|
|
@ -29,10 +31,12 @@ cp services/systemd/anvilsrv-user.service ~/.config/systemd/user/
|
|||
systemctl --user enable --now anvilsrv
|
||||
|
||||
# systemd system
|
||||
sudo cp services/systemd/anvilsrv.service /etc/systemd/system/ && sudo systemctl enable --now anvilsrv
|
||||
sudo cp services/systemd/anvilsrv.service /etc/systemd/system/
|
||||
sudo systemctl enable --now anvilsrv
|
||||
|
||||
# runit
|
||||
sudo cp -r services/runit /etc/sv/anvilsrv && sudo ln -s /etc/sv/anvilsrv /var/service/
|
||||
sudo cp -r services/runit /etc/sv/anvilsrv
|
||||
sudo ln -s /etc/sv/anvilsrv /var/service/
|
||||
```
|
||||
|
||||
## Usage
|
||||
|
|
@ -48,8 +52,8 @@ anvilsrv stop
|
|||
|
||||
**Namespaces:** Run multiple instances via `$NAMESPACE` (default: `/tmp/ns.$USER.:0`)
|
||||
```sh
|
||||
NAMESPACE=/tmp/ns.$USER.:1 anvilsrv start # :1
|
||||
NAMESPACE=/tmp/ns.$USER.:1 Assist # connect
|
||||
NAMESPACE=/tmp/ns.$USER.:1 anvilsrv start
|
||||
NAMESPACE=/tmp/ns.$USER.:1 Assist
|
||||
```
|
||||
|
||||
### Clients
|
||||
|
|
@ -60,7 +64,11 @@ NAMESPACE=/tmp/ns.$USER.:1 Assist # connect
|
|||
|
||||
**Web:** `anvilweb` (port :8080) — ⚠️ NO auth, localhost only — Remote: `ssh -L 8080:localhost:8080 user@remote`
|
||||
|
||||
**Acme:** Type `Assist` and middle-click — Tag: `Get Attach Stop Restart Kill Alias Context Daemon Inbox Archive` — Right-click ID for prompt window, 2-1 chord for fire-and-forget
|
||||
**Acme:** Type `Assist` and middle-click
|
||||
|
||||
**Tag commands:** `Get Attach Stop Restart Kill Alias Context Daemon Inbox Archive`
|
||||
|
||||
**Interaction:** Right-click ID for prompt window, 2-1 chord for fire-and-forget
|
||||
|
||||
| Command | Action |
|
||||
|---------|--------|
|
||||
|
|
@ -102,22 +110,54 @@ Available in sandbox YAML files (`~/.config/anvillm/`):
|
|||
| `{XDG_STATE_HOME}` | XDG state | `$XDG_STATE_HOME` or `~/.local/state` |
|
||||
| `{ANY_ENV_VAR}` | Environment variable | Any `$ENV_VAR` from the environment |
|
||||
|
||||
**Note:** Templates use `{VARNAME}` syntax. Any environment variable can be referenced.
|
||||
Templates use `{VARNAME}` syntax. Any environment variable can be referenced.
|
||||
|
||||
## Backends & Sandboxing
|
||||
|
||||
**Backends:** Claude (`npm install -g @anthropic-ai/claude-code`), Kiro ([kiro.dev](https://kiro.dev)), Ollama (local models via [ollie](https://github.com/lneely/ollie)) — Sandboxed with network access
|
||||
**Backends:** Claude (`npm install -g @anthropic-ai/claude-code`), Kiro ([kiro.dev](https://kiro.dev)), Ollama (local models via [ollie](https://github.com/lneely/ollie))
|
||||
|
||||
**Sandbox:** [landrun](https://github.com/zouuup/landrun) (always enabled) — Defaults: CWD/`/tmp`/config (rw), `/usr`/`/lib`/`/bin` (ro+exec), no network
|
||||
|
||||
**Config** (`~/.config/anvillm/`): Layered YAML files, most permissive wins:
|
||||
|
||||
<p align="center"><img src="docs/diagrams/sandbox-config.svg?v=2" width="250"></p>
|
||||
|
||||
- Default role: `roles/default.yaml`
|
||||
|
||||
```yaml
|
||||
network: {enabled: true, unrestricted: true}
|
||||
filesystem: {rw: ["{CWD}", "{HOME}/.npm"]}
|
||||
```
|
||||
|
||||
**Templates:** `{CWD}`, `{HOME}`, `{TMPDIR}`, `{XDG_*}` (see Configuration)
|
||||
|
||||
**Kernel requirements:** 5.13+ (Landlock v1), 6.7+ (v4), 6.10+ (v5 network)
|
||||
Set `best_effort: true` for unsandboxed fallback (⚠️ if no Landlock support)
|
||||
|
||||
**Session lifecycle:**
|
||||
|
||||
<p align="center"><img src="docs/diagrams/session-lifecycle.svg?v=2" width="500"></p>
|
||||
|
||||
State transitions: `idle` ↔ `running` cycle via CLI hooks (`userPromptSubmit` when user sends prompt, `stop` when agent finishes). Crash → `error` → auto-restart → `starting`. Note: any state can transition to `stopped` or `killed` (not shown); `stopped` can restart → `starting`.
|
||||
|
||||
**Self-healing:** Auto-restarts crashes every 5s (preserves context/alias/cwd), skips intentional stops
|
||||
|
||||
<p align="center"><img src="docs/diagrams/crash-recovery.svg?v=2" width="500"></p>
|
||||
|
||||
Restored sessions automatically resume the latest conversation (kiro: `-r`, claude: `-c`).
|
||||
|
||||
**Add backend:** Implement `CommandHandler`/`StateInspector` in `internal/backends/yourbackend.go`, register in `main.go`
|
||||
|
||||
### Ollama Backend
|
||||
|
||||
Run local LLMs via Ollama:
|
||||
Run local LLMs via Ollama.
|
||||
|
||||
**Requirements:**
|
||||
1. **Ollama:** Install from [ollama.com](https://ollama.com)
|
||||
```sh
|
||||
curl -fsSL https://ollama.com/install.sh | sh
|
||||
ollama serve # Start server (default: localhost:11434)
|
||||
ollama pull qwen2.5-coder:7b # Download a model
|
||||
ollama serve
|
||||
ollama pull qwen2.5-coder:7b
|
||||
```
|
||||
|
||||
2. **Ollie CLI:** Install from [github.com/lneely/ollie](https://github.com/lneely/ollie)
|
||||
|
|
@ -130,28 +170,11 @@ Run local LLMs via Ollama:
|
|||
echo 'new ollama /path/to/project' | 9p write agent/ctl
|
||||
```
|
||||
|
||||
**Configuration:** Set model via environment variable (default: `qwen3:8b`)
|
||||
**Configuration:** Set model via `ANVILLM_OLLAMA_MODEL` (default: `qwen3:8b`)
|
||||
```sh
|
||||
ANVILLM_OLLAMA_MODEL=llama3.2 echo 'new ollama /path' | 9p write agent/ctl
|
||||
```
|
||||
|
||||
**Add:** Implement `CommandHandler`/`StateInspector` in `internal/backends/yourbackend.go`, register in `main.go`
|
||||
|
||||
**Sandbox:** [landrun](https://github.com/zouuup/landrun) (cannot disable) — Defaults: CWD/`/tmp`/config (rw), `/usr`/`/lib`/`/bin` (ro+exec), no network
|
||||
|
||||
**Config** (`~/.config/anvillm/`): `global.yaml` → `backends/<name>.yaml` → `roles/<name>.yaml` (default: `roles/default.yaml`) → `tasks/<name>.yaml` — Most permissive wins
|
||||
|
||||
```yaml
|
||||
network: {enabled: true, unrestricted: true}
|
||||
filesystem: {rw: ["{CWD}", "{HOME}/.npm"]}
|
||||
```
|
||||
|
||||
Templates: `{CWD}`, `{HOME}`, `{TMPDIR}`, `{XDG_*}` (see Configuration) — Tip: Make `roles/default.yaml` permissive
|
||||
|
||||
**Kernel:** 5.13+ (Landlock v1), 6.7+ (v4), 6.10+ (v5 network) — Best-effort: `best_effort: true` (⚠️ unsandboxed if no Landlock)
|
||||
|
||||
**Self-healing:** Auto-restarts crashes every 5s (preserves context/alias/cwd) — Skips intentional stops
|
||||
|
||||
## 9P Filesystem
|
||||
|
||||
`$NAMESPACE/agent`:
|
||||
|
|
@ -188,6 +211,12 @@ agent/
|
|||
└── mail # Write messages (convenience)
|
||||
```
|
||||
|
||||
**Client Interactions:**
|
||||
|
||||
<p align="center"><img src="docs/diagrams/client-interactions.svg?v=2" width="400"></p>
|
||||
|
||||
Different clients interact with different parts of the filesystem: frontends read state, control files manage sessions, scripts consume events.
|
||||
|
||||
### Beads
|
||||
|
||||
Persistent task tracking via [beads](https://github.com/steveyegge/beads) — Dependency-aware graph
|
||||
|
|
@ -205,6 +234,14 @@ Config: `~/.beads/` (override: `ANVILLM_BEADS_PATH`) — Shared across namespace
|
|||
|
||||
### Events & Mailbox
|
||||
|
||||
**Mailbox Flow:**
|
||||
|
||||
|
||||
|
||||
Cross-backend communication: messages route between any participants (user, Claude agents, Kiro agents, Ollama agents) via the mailbox system.
|
||||
|
||||
**Examples:**
|
||||
|
||||
```sh
|
||||
# Events
|
||||
9p read agent/events # {"type":"state_change","session_id":"...","state":"running",...}
|
||||
|
|
@ -223,7 +260,6 @@ echo 'new claude /home/user/project' | 9p write agent/ctl
|
|||
|
||||
# List sessions (newest first)
|
||||
9p read agent/list
|
||||
# Output: <id> <backend> <state> <alias> <cwd>
|
||||
|
||||
# Get most recent session ID
|
||||
ID=$(9p read agent/list | head -1 | awk '{print $1}')
|
||||
|
|
@ -232,41 +268,43 @@ ID=$(9p read agent/list | head -1 | awk '{print $1}')
|
|||
echo '{"to":"'$ID'","type":"PROMPT_REQUEST","subject":"User prompt","body":"Hello"}' | 9p write user/mail
|
||||
|
||||
# Check state
|
||||
9p read agent/$ID/state # "running" or "idle"
|
||||
9p read agent/$ID/state
|
||||
|
||||
# Read response from inbox
|
||||
9p read user/inbox # Lists message files
|
||||
9p read user/inbox
|
||||
```
|
||||
|
||||
See `SECURITY.md`
|
||||
|
||||
## Workflows
|
||||
## Bot Templates
|
||||
|
||||
**Multi-Agent Communication:** Agents can communicate with each other **regardless of backend** (Claude, Kiro, etc.) via the built-in mailbox system. The `anvillm-communication` skill provides a convenient interface for agent messaging, enabling heterogeneous multi-agent workflows where different specialists use different backends while collaborating seamlessly.
|
||||
Bot templates provide pre-configured agents for common tasks. Templates are shell scripts that create sessions with specialized context and configuration.
|
||||
|
||||
**Skills:** `anvillm-skills list`, `anvillm-skills load anvillm-communication`
|
||||
|
||||
**Examples:** `./bot-templates/DevReview claude /path` (dev/reviewer loop), `./bot-templates/Planning kiro /path` (research/eng/editor)
|
||||
|
||||
**Pattern:**
|
||||
**Available Templates:**
|
||||
- `Taskmaster` — Converts project plans into dependency-aware task graphs (beads)
|
||||
- `Conductor` — Orchestrates parallel execution by spawning and coordinating specialized agents
|
||||
|
||||
**Usage:**
|
||||
```sh
|
||||
echo "new claude /project" | 9p write agent/ctl
|
||||
ID=$(9p read agent/list | tail -1 | awk '{print $1}')
|
||||
echo "You are..." | 9p write agent/$ID/context
|
||||
echo '{"to":"...","type":"...","subject":"...","body":"..."}' | 9p write agent/$ID/mail
|
||||
echo 'new "Task"' | 9p write agent/beads/ctl
|
||||
./bot-templates/Taskmaster kiro /path/to/project
|
||||
./bot-templates/Conductor kiro /path/to/project
|
||||
```
|
||||
|
||||
See `bot-templates/`, `kiro-cli/SKILLS_PROMPT.md`
|
||||
Templates set up context, roles, and initial configuration. Interact via any client (Assist, TUI, Emacs, Web) or directly via 9P mailbox.
|
||||
|
||||
## Automating Development
|
||||
**Creating Templates:** Shell scripts that write to `agent/ctl` and configure `context`, `role`, `tasks`. See existing templates for examples.
|
||||
|
||||
Automate development workflows with Taskmaster and Conductor bots. Input a project plan → Taskmaster creates tasks with dependencies → Conductor orchestrates parallel execution.
|
||||
## Autonomous Workflows
|
||||
|
||||
Automate workflows with Taskmaster and Conductor bots. Input project plan → Taskmaster creates tasks with dependencies → Conductor orchestrates parallel execution.
|
||||
|
||||
|
||||
|
||||
Conductor receives the top-level bead ID, analyzes dependencies, and spawns agents to work in parallel. Agents notify Conductor when blocked; Conductor signals them to resume when dependencies resolve.
|
||||
|
||||
### Usage
|
||||
|
||||
Examples below use raw 9P calls. These actions can also be performed from any front-end (Assist, TUI, Emacs, Web).
|
||||
Examples use raw 9P calls. These actions can also be performed from any front-end (Assist, TUI, Emacs, Web).
|
||||
|
||||
**1. Taskmaster: Create tasks from project plan**
|
||||
|
||||
|
|
@ -291,13 +329,13 @@ CONDUCTOR_ID=$(9p read agent/list | head -1 | awk '{print $1}')
|
|||
echo '{"to":"'$CONDUCTOR_ID'","type":"WORK_REQUEST","subject":"Execute","body":"Complete bead '$BEAD_ID'"}' | 9p write user/mail
|
||||
```
|
||||
|
||||
Conductor analyzes dependencies, spawns specialized bots (DevReview, Research, Docs), and delegates work in parallel.
|
||||
Conductor analyzes dependencies, spawns specialized bots, and delegates work in parallel.
|
||||
|
||||
**Monitoring:** Use any front-end (Assist, TUI, Emacs, Web). For custom integrations, see `EVENTS.md`.
|
||||
|
||||
## MCP Integration
|
||||
|
||||
`anvilmcp` exposes AnviLLM via Model Context Protocol for Claude Desktop, Cline, etc
|
||||
`anvilmcp` exposes AnviLLM via Model Context Protocol for Claude Desktop, Cline, etc.
|
||||
|
||||
**Install:**
|
||||
```sh
|
||||
|
|
@ -306,7 +344,7 @@ Conductor analyzes dependencies, spawns specialized bots (DevReview, Research, D
|
|||
|
||||
**Tools:** `read_inbox`, `send_message`, `list_sessions`, `set_state`
|
||||
|
||||
Enables LLM clients to control AnviLLM sessions and communicate with agents
|
||||
Enables LLM clients to discover agents and communicate via the mailbox system.
|
||||
|
||||
## Troubleshooting
|
||||
|
||||
|
|
|
|||
|
|
@ -1,122 +0,0 @@
|
|||
#!/bin/bash
|
||||
# DevReview - Developer-Reviewer Workflow
|
||||
# Usage: ./scripts/DevReview <backend> <workdir>
|
||||
|
||||
set -e
|
||||
|
||||
if [ $# -ne 2 ]; then
|
||||
echo "Usage: $0 <backend> <workdir>"
|
||||
echo " backend: claude, kiro-cli"
|
||||
exit 1
|
||||
fi
|
||||
|
||||
BACKEND="$1"
|
||||
PROJECT_PATH="$2"
|
||||
PROJECT_NAME=$(basename "$PROJECT_PATH")
|
||||
|
||||
# Generate a random short identifier
|
||||
RANDOM_WORD=$(shuf -n1 -e alpha beta gamma delta zeta theta kappa sigma omega lambda)
|
||||
RANDOM_NUM=$(printf "%02d" $((RANDOM % 100)))
|
||||
ALIAS_NAME="${RANDOM_WORD}${RANDOM_NUM}"
|
||||
|
||||
echo "Setting up Developer-Reviewer workflow for: $PROJECT_NAME"
|
||||
echo "Project path: $PROJECT_PATH"
|
||||
echo "Session alias: $ALIAS_NAME"
|
||||
|
||||
NAMESPACE=$(namespace)
|
||||
if [ -z "$NAMESPACE" ]; then
|
||||
echo "Error: No namespace found. Is anvillm running?"
|
||||
exit 1
|
||||
fi
|
||||
|
||||
AGENT_MOUNT="agent"
|
||||
|
||||
echo "Creating developer session..."
|
||||
before=$(9p read $AGENT_MOUNT/list | awk '{print $1}' | sort)
|
||||
echo "new $BACKEND $PROJECT_PATH role=developer tasks=git,editor,go,rust,python,nodejs,docker,podman,maven,gradle,plan9,gpg,github-cli,agent-q" | 9p write $AGENT_MOUNT/ctl
|
||||
after=$(9p read $AGENT_MOUNT/list | awk '{print $1}' | sort)
|
||||
DEV_ID=$(comm -13 <(echo "$before") <(echo "$after") | head -1)
|
||||
echo "Developer session ID: $DEV_ID"
|
||||
|
||||
echo "Creating reviewer session..."
|
||||
before=$(9p read $AGENT_MOUNT/list | awk '{print $1}' | sort)
|
||||
echo "new $BACKEND $PROJECT_PATH role=reviewer tasks=git,github-cli,plan9,agent-q" | 9p write $AGENT_MOUNT/ctl
|
||||
after=$(9p read $AGENT_MOUNT/list | awk '{print $1}' | sort)
|
||||
REVIEWER_ID=$(comm -13 <(echo "$before") <(echo "$after") | head -1)
|
||||
echo "Reviewer session ID: $REVIEWER_ID"
|
||||
|
||||
# Set aliases
|
||||
echo "Setting aliases..."
|
||||
echo "${ALIAS_NAME}-dev" | 9p write $AGENT_MOUNT/$DEV_ID/alias
|
||||
echo "${ALIAS_NAME}-reviewer" | 9p write $AGENT_MOUNT/$REVIEWER_ID/alias
|
||||
|
||||
# Set developer context
|
||||
cat <<EOF | 9p write $AGENT_MOUNT/$DEV_ID/context
|
||||
# DEVELOPER AGENT ROLE
|
||||
|
||||
You are the developer agent in a two-agent AnviLLM workflow. Your peer is the reviewer agent.
|
||||
|
||||
## Your workflow:
|
||||
|
||||
1. When given a task, implement it fully
|
||||
2. After implementation, stage all changes using git
|
||||
3. Send REVIEW_REQUEST to the reviewer (find by alias containing "reviewer")
|
||||
4. When you receive REVIEW_RESPONSE with "LGTM", send PROMPT_RESPONSE to user summarizing what was done, then you're done
|
||||
5. When you receive REVIEW_RESPONSE with suggested changes, implement them and repeat from step 2
|
||||
|
||||
## MANDATORY: Status Update to User
|
||||
|
||||
When your task is complete (after receiving LGTM), you MUST send PROMPT_RESPONSE to user with a summary of what was accomplished and key changes made.
|
||||
|
||||
**Never finish without sending this status update to the user.**
|
||||
|
||||
Remember: You are working collaboratively with the reviewer to ensure code quality.
|
||||
|
||||
For communication with other agents or the user, load the anvillm-communication skill.
|
||||
EOF
|
||||
|
||||
# Set reviewer context
|
||||
cat <<EOF | 9p write $AGENT_MOUNT/$REVIEWER_ID/context
|
||||
# REVIEWER AGENT ROLE
|
||||
|
||||
You are the code reviewer agent in a two-agent AnviLLM workflow. Your peer is the developer agent.
|
||||
|
||||
## Your workflow:
|
||||
|
||||
1. When you receive a REVIEW_REQUEST, review the staged changes
|
||||
2. Use git commands to examine what's been staged:
|
||||
- git diff --cached
|
||||
- Look for bugs, code quality issues, security concerns
|
||||
3. Send REVIEW_RESPONSE to the developer (find by alias containing "dev"):
|
||||
- If code is good: body should be "LGTM"
|
||||
- If changes needed: body should contain detailed feedback
|
||||
|
||||
## Review Criteria:
|
||||
|
||||
- Code correctness
|
||||
- Security (no SQL injection, XSS, command injection, etc.)
|
||||
- Code style and readability
|
||||
- Test coverage
|
||||
- Error handling
|
||||
|
||||
Remember: Be thorough but constructive in your reviews. The goal is high-quality code.
|
||||
|
||||
For communication with other agents, load the anvillm-communication skill.
|
||||
EOF
|
||||
|
||||
echo ""
|
||||
echo "✓ Developer-Reviewer workflow setup complete!"
|
||||
echo ""
|
||||
echo "Sessions created:"
|
||||
echo " Developer: $DEV_ID (alias: ${ALIAS_NAME}-dev)"
|
||||
echo " Reviewer: $REVIEWER_ID (alias: ${ALIAS_NAME}-reviewer)"
|
||||
echo ""
|
||||
echo "To start development:"
|
||||
echo " 1. Open the developer agent and give it a task"
|
||||
echo " 2. The developer will implement, then automatically request review"
|
||||
echo " 3. The reviewer will provide feedback"
|
||||
echo " 4. The cycle continues until the reviewer gives LGTM"
|
||||
echo ""
|
||||
echo "To interact with the agents:"
|
||||
echo " - Open developer: echo open $DEV_ID | 9p write $AGENT_MOUNT/ctl"
|
||||
echo " - Open reviewer: echo open $REVIEWER_ID | 9p write $AGENT_MOUNT/ctl"
|
||||
|
|
@ -1,172 +0,0 @@
|
|||
#!/bin/bash
|
||||
# DevReview - Developer-Reviewer Workflow
|
||||
# Usage: ./scripts/DevReview <backend> <workdir>
|
||||
|
||||
set -e
|
||||
|
||||
if [ $# -ne 2 ]; then
|
||||
echo "Usage: $0 <backend> <workdir>"
|
||||
echo " backend: claude, kiro-cli"
|
||||
exit 1
|
||||
fi
|
||||
|
||||
BACKEND="$1"
|
||||
PROJECT_PATH="$2"
|
||||
PROJECT_NAME=$(basename "$PROJECT_PATH")
|
||||
|
||||
# Generate a random short identifier
|
||||
RANDOM_WORD=$(shuf -n1 -e alpha beta gamma delta zeta theta kappa sigma omega lambda)
|
||||
RANDOM_NUM=$(printf "%02d" $((RANDOM % 100)))
|
||||
ALIAS_NAME="${RANDOM_WORD}${RANDOM_NUM}"
|
||||
|
||||
echo "Setting up Developer-Reviewer workflow for: $PROJECT_NAME"
|
||||
echo "Project path: $PROJECT_PATH"
|
||||
echo "Session alias: $ALIAS_NAME"
|
||||
|
||||
NAMESPACE=$(namespace)
|
||||
if [ -z "$NAMESPACE" ]; then
|
||||
echo "Error: No namespace found. Is anvillm running?"
|
||||
exit 1
|
||||
fi
|
||||
|
||||
AGENT_MOUNT="agent"
|
||||
|
||||
echo "Creating developer session..."
|
||||
before=$(9p read $AGENT_MOUNT/list | awk '{print $1}' | sort)
|
||||
echo "new $BACKEND $PROJECT_PATH role=developer tasks=git,editor,go,rust,python,nodejs,docker,podman,maven,gradle,plan9,gpg,github-cli,agent-q" | 9p write $AGENT_MOUNT/ctl
|
||||
after=$(9p read $AGENT_MOUNT/list | awk '{print $1}' | sort)
|
||||
DEV_ID=$(comm -13 <(echo "$before") <(echo "$after") | head -1)
|
||||
echo "Developer session ID: $DEV_ID"
|
||||
|
||||
echo "Creating reviewer session..."
|
||||
before=$(9p read $AGENT_MOUNT/list | awk '{print $1}' | sort)
|
||||
echo "new $BACKEND $PROJECT_PATH role=reviewer tasks=git,github-cli,plan9,agent-q" | 9p write $AGENT_MOUNT/ctl
|
||||
after=$(9p read $AGENT_MOUNT/list | awk '{print $1}' | sort)
|
||||
REVIEWER_ID=$(comm -13 <(echo "$before") <(echo "$after") | head -1)
|
||||
echo "Reviewer session ID: $REVIEWER_ID"
|
||||
|
||||
echo "Creating QA session..."
|
||||
before=$(9p read $AGENT_MOUNT/list | awk '{print $1}' | sort)
|
||||
echo "new $BACKEND $PROJECT_PATH role=tester tasks=git,github-cli,plan9,agent-q" | 9p write $AGENT_MOUNT/ctl
|
||||
after=$(9p read $AGENT_MOUNT/list | awk '{print $1}' | sort)
|
||||
QA_ID=$(comm -13 <(echo "$before") <(echo "$after") | head -1)
|
||||
echo "QA session ID: $QA_ID"
|
||||
|
||||
# Set aliases
|
||||
echo "Setting aliases..."
|
||||
echo "${ALIAS_NAME}-dev" | 9p write $AGENT_MOUNT/$DEV_ID/alias
|
||||
echo "${ALIAS_NAME}-reviewer" | 9p write $AGENT_MOUNT/$REVIEWER_ID/alias
|
||||
echo "${ALIAS_NAME}-qa" | 9p write $AGENT_MOUNT/$QA_ID/alias
|
||||
|
||||
# Set developer context
|
||||
cat <<EOF | 9p write $AGENT_MOUNT/$DEV_ID/context
|
||||
# DEVELOPER AGENT ROLE
|
||||
|
||||
You are the developer agent in a three-agent AnviLLM workflow. Your peers are the reviewer agent and the QA agent.
|
||||
|
||||
## Your workflow:
|
||||
|
||||
1. When given a task, implement it fully
|
||||
2. After implementation, stage all changes using git
|
||||
3. Send REVIEW_REQUEST to the reviewer (find by alias containing "reviewer")
|
||||
4. When you receive REVIEW_RESPONSE with suggested changes, implement them and repeat from step 2
|
||||
5. When you receive REVIEW_RESPONSE with "LGTM":
|
||||
- Send APPROVAL_REQUEST to QA (find by alias containing "qa")
|
||||
- Include: test spec with APIs, test commands, expected behavior
|
||||
6. When you receive APPROVAL_RESPONSE with issues, fix them and repeat from step 2
|
||||
7. When you receive APPROVAL_RESPONSE confirming all tests pass, you're done
|
||||
|
||||
Remember: You are working collaboratively with the reviewer and QA to ensure code quality and correctness.
|
||||
|
||||
For communication with other agents, load the anvillm-communication skill.
|
||||
EOF
|
||||
|
||||
# Set reviewer context
|
||||
cat <<EOF | 9p write $AGENT_MOUNT/$REVIEWER_ID/context
|
||||
# REVIEWER AGENT ROLE
|
||||
|
||||
You are the code reviewer agent in a three-agent AnviLLM workflow. Your peers are the developer agent and the QA agent.
|
||||
|
||||
## Your workflow:
|
||||
|
||||
1. When you receive a REVIEW_REQUEST, review the staged changes
|
||||
2. Use git commands to examine what's been staged:
|
||||
- git diff --cached
|
||||
- Look for bugs, code quality issues, security concerns
|
||||
3. Send REVIEW_RESPONSE to the developer (find by alias containing "dev"):
|
||||
- If code is good: body should be "LGTM"
|
||||
- If changes needed: body should contain detailed feedback
|
||||
|
||||
## Review Criteria:
|
||||
|
||||
- Code correctness
|
||||
- Security (no SQL injection, XSS, command injection, etc.)
|
||||
- Code style and readability
|
||||
- Test coverage
|
||||
- Error handling
|
||||
|
||||
Remember: Be thorough but constructive in your reviews. The goal is high-quality code.
|
||||
|
||||
For communication with other agents, load the anvillm-communication skill.
|
||||
EOF
|
||||
|
||||
# Set QA context
|
||||
cat <<EOF | 9p write $AGENT_MOUNT/$QA_ID/context
|
||||
# QA AGENT ROLE
|
||||
|
||||
You are the QA agent in a three-agent AnviLLM workflow. Your peers are the developer agent and the reviewer agent.
|
||||
|
||||
## Your workflow:
|
||||
|
||||
1. When you receive an APPROVAL_REQUEST from the developer, it will include:
|
||||
- A description of what was implemented
|
||||
- How to test it (commands, API endpoints, etc.)
|
||||
- Expected behavior
|
||||
2. Execute the specified tests:
|
||||
- Run any test commands provided
|
||||
- Test public APIs if applicable (e.g., 9P filesystem interface, CLI commands, HTTP endpoints)
|
||||
- Verify expected behavior matches actual behavior
|
||||
- Check for edge cases and error handling
|
||||
3. Test for functional correctness and quality:
|
||||
- Does it work as specified?
|
||||
- Are there any bugs or unexpected behaviors?
|
||||
- Is error handling appropriate?
|
||||
- Are edge cases handled?
|
||||
- Is the user experience good?
|
||||
4. Send APPROVAL_RESPONSE to the developer (find by alias containing "dev"):
|
||||
- If all tests pass: confirm tests passed
|
||||
- If issues found: provide detailed failure report
|
||||
|
||||
## Testing Focus:
|
||||
|
||||
- Functional correctness (does it do what it should?)
|
||||
- Quality (is it done well?)
|
||||
- Edge cases and error conditions
|
||||
- User experience
|
||||
- Integration with existing systems
|
||||
|
||||
Remember: Your job is to ensure the software works correctly and meets quality standards. Be thorough but practical.
|
||||
|
||||
For communication with other agents, load the anvillm-communication skill.
|
||||
EOF
|
||||
|
||||
echo ""
|
||||
echo "✓ Developer-Reviewer-QA workflow setup complete!"
|
||||
echo ""
|
||||
echo "Sessions created:"
|
||||
echo " Developer: $DEV_ID (alias: ${ALIAS_NAME}-dev)"
|
||||
echo " Reviewer: $REVIEWER_ID (alias: ${ALIAS_NAME}-reviewer)"
|
||||
echo " QA: $QA_ID (alias: ${ALIAS_NAME}-qa)"
|
||||
echo ""
|
||||
echo "To start development:"
|
||||
echo " 1. Open the developer agent and give it a task"
|
||||
echo " 2. The developer will implement, then automatically request review"
|
||||
echo " 3. The reviewer will provide feedback or LGTM"
|
||||
echo " 4. After LGTM, the developer sends test request to QA"
|
||||
echo " 5. QA tests the implementation and reports results"
|
||||
echo " 6. The cycle continues until QA confirms all tests pass"
|
||||
echo ""
|
||||
echo "To interact with the agents:"
|
||||
echo " - Open developer: echo open $DEV_ID | 9p write $AGENT_MOUNT/ctl"
|
||||
echo " - Open reviewer: echo open $REVIEWER_ID | 9p write $AGENT_MOUNT/ctl"
|
||||
echo " - Open QA: echo open $QA_ID | 9p write $AGENT_MOUNT/ctl"
|
||||
|
|
@ -1,68 +0,0 @@
|
|||
#!/bin/bash
|
||||
# Docs - Documentation workflow (writer + techeditor)
|
||||
# Usage: ./bot-templates/Docs <backend> <workdir>
|
||||
|
||||
set -e
|
||||
|
||||
if [ $# -ne 2 ]; then
|
||||
echo "Usage: $0 <backend> <workdir>"
|
||||
echo " backend: claude, kiro-cli"
|
||||
exit 1
|
||||
fi
|
||||
|
||||
BACKEND="$1"
|
||||
DOC_PATH="$2"
|
||||
|
||||
RANDOM_WORD=$(shuf -n1 -e alpha beta gamma delta zeta theta kappa sigma omega lambda)
|
||||
RANDOM_NUM=$(printf "%02d" $((RANDOM % 100)))
|
||||
ALIAS_PREFIX="${RANDOM_WORD}${RANDOM_NUM}"
|
||||
|
||||
NAMESPACE=$(namespace)
|
||||
if [ -z "$NAMESPACE" ]; then
|
||||
echo "Error: No namespace found. Is anvillm running?"
|
||||
exit 1
|
||||
fi
|
||||
|
||||
AGENT_MOUNT="agent"
|
||||
|
||||
echo "Creating writer session..."
|
||||
before=$(9p read $AGENT_MOUNT/list | awk '{print $1}' | sort)
|
||||
echo "new $BACKEND $DOC_PATH role=writer tasks=git,plan9,agent-q" | 9p write $AGENT_MOUNT/ctl
|
||||
after=$(9p read $AGENT_MOUNT/list | awk '{print $1}' | sort)
|
||||
WRITER_ID=$(comm -13 <(echo "$before") <(echo "$after") | head -1)
|
||||
|
||||
echo "Creating techeditor session..."
|
||||
before=$(9p read $AGENT_MOUNT/list | awk '{print $1}' | sort)
|
||||
echo "new $BACKEND $DOC_PATH role=techeditor tasks=git,plan9,agent-q" | 9p write $AGENT_MOUNT/ctl
|
||||
after=$(9p read $AGENT_MOUNT/list | awk '{print $1}' | sort)
|
||||
EDITOR_ID=$(comm -13 <(echo "$before") <(echo "$after") | head -1)
|
||||
|
||||
echo "${ALIAS_PREFIX}-writer" | 9p write $AGENT_MOUNT/$WRITER_ID/alias
|
||||
echo "${ALIAS_PREFIX}-techeditor" | 9p write $AGENT_MOUNT/$EDITOR_ID/alias
|
||||
|
||||
cat <<EOF | 9p write $AGENT_MOUNT/$WRITER_ID/context
|
||||
Your job is to write and update documentation in $DOC_PATH.
|
||||
|
||||
You are a writer agent in a two-agent AnviLLM workflow. Your peer is the techeditor.
|
||||
|
||||
After completing work, send APPROVAL_REQUEST to the techeditor for review. You will receive APPROVAL_RESPONSE with feedback or approval.
|
||||
|
||||
For communication with other agents, load the anvillm-communication skill.
|
||||
EOF
|
||||
|
||||
cat <<EOF | 9p write $AGENT_MOUNT/$EDITOR_ID/context
|
||||
You are a professional technical editor in a two-agent AnviLLM workflow. Your peer is the writer. Ensure documentation in $DOC_PATH is high quality, readable, and at the right detail level.
|
||||
|
||||
When you receive an APPROVAL_REQUEST from the writer, review the documentation. Send APPROVAL_RESPONSE with either approval or recommended changes.
|
||||
|
||||
For communication with other agents, load the anvillm-communication skill.
|
||||
EOF
|
||||
|
||||
echo ""
|
||||
echo "✓ Documentation workflow ready ($ALIAS_PREFIX)"
|
||||
echo ""
|
||||
echo "Sessions:"
|
||||
echo " ${ALIAS_PREFIX}-writer: $WRITER_ID"
|
||||
echo " ${ALIAS_PREFIX}-techeditor: $EDITOR_ID"
|
||||
echo ""
|
||||
echo "Give the writer a documentation task, or techeditor a document to review."
|
||||
|
|
@ -1,101 +0,0 @@
|
|||
#!/bin/bash
|
||||
# Planning - Research/Architect/Techeditor workflow
|
||||
# Usage: ./scripts/Planning <backend> <workdir>
|
||||
|
||||
set -e
|
||||
|
||||
if [ $# -ne 2 ]; then
|
||||
echo "Usage: $0 <backend> <workdir>"
|
||||
echo " backend: claude, kiro-cli"
|
||||
exit 1
|
||||
fi
|
||||
|
||||
BACKEND="$1"
|
||||
DOC_PATH="$2"
|
||||
|
||||
# Generate random alias prefix
|
||||
RANDOM_WORD=$(shuf -n1 -e alpha beta gamma delta zeta theta kappa sigma omega lambda)
|
||||
RANDOM_NUM=$(printf "%02d" $((RANDOM % 100)))
|
||||
ALIAS_PREFIX="${RANDOM_WORD}${RANDOM_NUM}"
|
||||
|
||||
NAMESPACE=$(namespace)
|
||||
if [ -z "$NAMESPACE" ]; then
|
||||
echo "Error: No namespace found. Is anvillm running?"
|
||||
exit 1
|
||||
fi
|
||||
|
||||
AGENT_MOUNT="agent"
|
||||
|
||||
echo "Creating research session..."
|
||||
before=$(9p read $AGENT_MOUNT/list | awk '{print $1}' | sort)
|
||||
echo "new $BACKEND $DOC_PATH role=research tasks=plan9,agent-q" | 9p write $AGENT_MOUNT/ctl
|
||||
after=$(9p read $AGENT_MOUNT/list | awk '{print $1}' | sort)
|
||||
RESEARCH_ID=$(comm -13 <(echo "$before") <(echo "$after") | head -1)
|
||||
echo " ID: $RESEARCH_ID"
|
||||
|
||||
echo "Creating architect session..."
|
||||
before=$(9p read $AGENT_MOUNT/list | awk '{print $1}' | sort)
|
||||
echo "new $BACKEND $DOC_PATH role=architect tasks=git,plan9,agent-q" | 9p write $AGENT_MOUNT/ctl
|
||||
after=$(9p read $AGENT_MOUNT/list | awk '{print $1}' | sort)
|
||||
ARCH_ID=$(comm -13 <(echo "$before") <(echo "$after") | head -1)
|
||||
echo " ID: $ARCH_ID"
|
||||
|
||||
echo "Creating techeditor session..."
|
||||
before=$(9p read $AGENT_MOUNT/list | awk '{print $1}' | sort)
|
||||
echo "new $BACKEND $DOC_PATH role=techeditor tasks=git,plan9,agent-q" | 9p write $AGENT_MOUNT/ctl
|
||||
after=$(9p read $AGENT_MOUNT/list | awk '{print $1}' | sort)
|
||||
EDITOR_ID=$(comm -13 <(echo "$before") <(echo "$after") | head -1)
|
||||
echo " ID: $EDITOR_ID"
|
||||
|
||||
# Set aliases
|
||||
echo "${ALIAS_PREFIX}-research" | 9p write $AGENT_MOUNT/$RESEARCH_ID/alias
|
||||
echo "${ALIAS_PREFIX}-architect" | 9p write $AGENT_MOUNT/$ARCH_ID/alias
|
||||
echo "${ALIAS_PREFIX}-techeditor" | 9p write $AGENT_MOUNT/$EDITOR_ID/alias
|
||||
|
||||
# Research context
|
||||
cat <<EOF | 9p write $AGENT_MOUNT/$RESEARCH_ID/context
|
||||
Your job is to provide answers to questions by researching existing knowledge base documents, the web, and the code base.
|
||||
|
||||
You are a research agent in a three-agent AnviLLM workflow. Your peers are the architect and techeditor.
|
||||
|
||||
- Input: knowledge base (agent-kb) and relevant code
|
||||
- Output: concrete human- and LLM-readable knowledge about the system
|
||||
- Prefer updating existing kb documents over creating new ones
|
||||
- Only create new KB documents if the knowledge discovered is significant and reusable
|
||||
|
||||
When you receive a QUERY_REQUEST from the architect, research and respond with QUERY_RESPONSE.
|
||||
|
||||
For communication with other agents, load the anvillm-communication skill.
|
||||
EOF
|
||||
|
||||
# Architect context
|
||||
cat <<EOF | 9p write $AGENT_MOUNT/$ARCH_ID/context
|
||||
Your job is to update design documentation in $DOC_PATH. Follow agent-kb conventions. Search the knowledge base for relevant documents.
|
||||
|
||||
You are an architect agent in a three-agent AnviLLM workflow. Your peers are the research agent and techeditor.
|
||||
|
||||
If documentation is insufficient, send QUERY_REQUEST to the research bot and wait for QUERY_RESPONSE.
|
||||
|
||||
After completing work, send APPROVAL_REQUEST to the techeditor for review. You will receive APPROVAL_RESPONSE with feedback or approval.
|
||||
|
||||
For communication with other agents, load the anvillm-communication skill.
|
||||
EOF
|
||||
|
||||
# Tech-editor context
|
||||
cat <<EOF | 9p write $AGENT_MOUNT/$EDITOR_ID/context
|
||||
You are a professional technical writer in a three-agent AnviLLM workflow. Your peers are the research agent and architect. Ensure documentation in $DOC_PATH is high quality, readable, and at the right detail level for software engineers.
|
||||
|
||||
When you receive an APPROVAL_REQUEST from the architect, review the documentation. Send APPROVAL_RESPONSE with either approval or recommended changes.
|
||||
|
||||
For communication with other agents, load the anvillm-communication skill.
|
||||
EOF
|
||||
|
||||
echo ""
|
||||
echo "✓ Planning workflow ready ($ALIAS_PREFIX)"
|
||||
echo ""
|
||||
echo "Sessions:"
|
||||
echo " ${ALIAS_PREFIX}-research: $RESEARCH_ID"
|
||||
echo " ${ALIAS_PREFIX}-architect: $ARCH_ID"
|
||||
echo " ${ALIAS_PREFIX}-techeditor: $EDITOR_ID"
|
||||
echo ""
|
||||
echo "Start by giving the architect a task, or techeditor a document to review."
|
||||
|
|
@ -1,57 +0,0 @@
|
|||
#!/bin/bash
|
||||
# Research - Single research agent workflow
|
||||
# Usage: ./bot-templates/Research <backend> <workdir>
|
||||
|
||||
set -e
|
||||
|
||||
if [ $# -ne 2 ]; then
|
||||
echo "Usage: $0 <backend> <workdir>"
|
||||
echo " backend: claude, kiro-cli"
|
||||
exit 1
|
||||
fi
|
||||
|
||||
BACKEND="$1"
|
||||
WORKDIR="$2"
|
||||
|
||||
RANDOM_WORD=$(shuf -n1 -e alpha beta gamma delta zeta theta kappa sigma omega lambda)
|
||||
RANDOM_NUM=$(printf "%02d" $((RANDOM % 100)))
|
||||
ALIAS_NAME="${RANDOM_WORD}${RANDOM_NUM}-research"
|
||||
|
||||
NAMESPACE=$(namespace)
|
||||
if [ -z "$NAMESPACE" ]; then
|
||||
echo "Error: No namespace found. Is anvillm running?"
|
||||
exit 1
|
||||
fi
|
||||
|
||||
AGENT_MOUNT="agent"
|
||||
|
||||
echo "Creating research session..."
|
||||
before=$(9p read $AGENT_MOUNT/list | awk '{print $1}' | sort)
|
||||
echo "new $BACKEND $WORKDIR role=research tasks=plan9,agent-q" | 9p write $AGENT_MOUNT/ctl
|
||||
after=$(9p read $AGENT_MOUNT/list | awk '{print $1}' | sort)
|
||||
RESEARCH_ID=$(comm -13 <(echo "$before") <(echo "$after") | head -1)
|
||||
|
||||
echo "$ALIAS_NAME" | 9p write $AGENT_MOUNT/$RESEARCH_ID/alias
|
||||
|
||||
cat <<EOF | 9p write $AGENT_MOUNT/$RESEARCH_ID/context
|
||||
Your job is to provide answers to questions by researching existing knowledge base documents, the web, and the code base.
|
||||
|
||||
You are a research agent in a single-agent AnviLLM workflow.
|
||||
|
||||
- Search existing knowledge base documents for relevant information
|
||||
- Analyze code and documentation in $WORKDIR
|
||||
- Create or update KB documents with your findings
|
||||
- Prefer updating existing documents over creating new ones
|
||||
- Only create new KB documents for significant, reusable knowledge
|
||||
|
||||
Focus on providing concrete, actionable information that helps understand the system.
|
||||
|
||||
For communication with other agents or the user, load the anvillm-communication skill.
|
||||
EOF
|
||||
|
||||
echo ""
|
||||
echo "✓ Research workflow ready"
|
||||
echo ""
|
||||
echo "Session: $RESEARCH_ID (alias: $ALIAS_NAME)"
|
||||
echo ""
|
||||
echo "Give the research agent a topic to investigate."
|
||||
|
|
@ -0,0 +1,30 @@
|
|||
graph TB
|
||||
subgraph Clients
|
||||
Assist["Assist<br/>(Acme)"]
|
||||
TUI["TUI"]
|
||||
Emacs["Emacs"]
|
||||
Web["Web"]
|
||||
end
|
||||
|
||||
subgraph Core["Core Services"]
|
||||
anvilsrv["anvilsrv<br/>(9P daemon)"]
|
||||
anvilmcp["anvilmcp<br/>(MCP server)"]
|
||||
anvilwebgw["anvilwebgw<br/>(API proxy)"]
|
||||
end
|
||||
|
||||
subgraph Backends
|
||||
Claude["Claude"]
|
||||
Kiro["Kiro"]
|
||||
ollie["ollie"]
|
||||
Ollama["Ollama"]
|
||||
end
|
||||
|
||||
Assist -->|9P| anvilsrv
|
||||
TUI -->|9P| anvilsrv
|
||||
Emacs -->|9P| anvilsrv
|
||||
Web -->|9P| anvilsrv
|
||||
|
||||
anvilsrv --> Claude
|
||||
anvilsrv --> Kiro
|
||||
anvilsrv --> ollie
|
||||
ollie --> Ollama
|
||||
File diff suppressed because one or more lines are too long
|
After Width: | Height: | Size: 20 KiB |
|
|
@ -0,0 +1,36 @@
|
|||
sequenceDiagram
|
||||
participant User
|
||||
participant Taskmaster
|
||||
participant Beads
|
||||
participant Conductor
|
||||
participant Bot1 as Bot 1
|
||||
participant Bot2 as Bot 2
|
||||
participant Bot3 as Bot 3
|
||||
|
||||
User->>Taskmaster: Project plan
|
||||
Taskmaster->>Beads: Create task graph
|
||||
Beads-->>Taskmaster: Top-level bead ID
|
||||
Taskmaster-->>User: Tasks created
|
||||
|
||||
User->>Conductor: Top-level bead ID
|
||||
Conductor->>Beads: Analyze dependencies
|
||||
|
||||
Conductor->>Bot1: Assign task (no blockers)
|
||||
Conductor->>Bot2: Assign task (no blockers)
|
||||
|
||||
Bot1->>Bot1: Work on task
|
||||
Bot1-->>Conductor: Task complete
|
||||
|
||||
Conductor->>Beads: Check dependencies
|
||||
Conductor->>Bot3: Assign task (blockers resolved)
|
||||
|
||||
Bot2->>Bot2: Work on task
|
||||
Bot2-->>Conductor: Task complete
|
||||
|
||||
Bot3->>Bot3: Work on task
|
||||
Bot3-->>Conductor: Task complete
|
||||
|
||||
Conductor->>Beads: Check all child beads
|
||||
Beads-->>Conductor: All complete
|
||||
Conductor->>Beads: Mark top-level complete
|
||||
Conductor-->>User: All work done
|
||||
File diff suppressed because one or more lines are too long
|
After Width: | Height: | Size: 30 KiB |
|
|
@ -0,0 +1,9 @@
|
|||
graph TB
|
||||
Assist["Assist"]
|
||||
TUI["TUI"]
|
||||
Script["Script"]
|
||||
FS["agent/<br/>(9P filesystem)"]
|
||||
|
||||
Assist -->|9p read<br/>agent/list| FS
|
||||
TUI -->|9p write<br/>agent/ctl| FS
|
||||
Script -->|9p read<br/>agent/events| FS
|
||||
File diff suppressed because one or more lines are too long
|
After Width: | Height: | Size: 10 KiB |
|
|
@ -0,0 +1,10 @@
|
|||
graph LR
|
||||
running1[running]
|
||||
error[error]
|
||||
running2[running]
|
||||
state[context<br/>alias<br/>cwd<br/>conversation]
|
||||
|
||||
running1 -->|crash| error
|
||||
error -->|preserves| state
|
||||
state -->|restored +<br/>auto-resume| running2
|
||||
error -->|5s wait| running2
|
||||
File diff suppressed because one or more lines are too long
|
After Width: | Height: | Size: 11 KiB |
|
|
@ -0,0 +1,9 @@
|
|||
graph TB
|
||||
events["agent/events<br/>(event stream)"]
|
||||
notif["Notification"]
|
||||
logger["Logger<br/>(audit)"]
|
||||
etc["etc."]
|
||||
|
||||
events --> notif
|
||||
events --> logger
|
||||
events --> etc
|
||||
File diff suppressed because one or more lines are too long
|
After Width: | Height: | Size: 9.9 KiB |
|
|
@ -0,0 +1,17 @@
|
|||
sequenceDiagram
|
||||
participant User
|
||||
participant Mailbox as Mailbox System
|
||||
participant AgentA as Agent A (Claude)
|
||||
participant AgentB as Agent B (Kiro)
|
||||
|
||||
User->>Mailbox: write user/mail<br/>{"to":"agent-a",...}
|
||||
Mailbox->>AgentA: agent-a/inbox
|
||||
|
||||
AgentA->>Mailbox: write agent/agent-a/mail<br/>{"to":"agent-b",...}
|
||||
Mailbox->>AgentB: agent-b/inbox
|
||||
|
||||
AgentB->>Mailbox: write agent/agent-b/mail<br/>{"to":"agent-a",...}
|
||||
Mailbox->>AgentA: agent-a/inbox
|
||||
|
||||
AgentB->>Mailbox: write agent/agent-b/mail<br/>{"to":"user",...}
|
||||
Mailbox->>User: user/inbox
|
||||
File diff suppressed because one or more lines are too long
|
After Width: | Height: | Size: 24 KiB |
|
|
@ -0,0 +1,11 @@
|
|||
graph TB
|
||||
global["global.yaml<br/>(Base configuration)"]
|
||||
backends["backends/*.yaml<br/>(Backend-specific)"]
|
||||
roles["roles/*.yaml<br/>(Role-specific)<br/>default: roles/default.yaml"]
|
||||
tasks["tasks/*.yaml<br/>(Task-specific)"]
|
||||
final["Final Config<br/>(most permissive wins)"]
|
||||
|
||||
global -->|override| backends
|
||||
backends -->|override| roles
|
||||
roles -->|override| tasks
|
||||
tasks --> final
|
||||
File diff suppressed because one or more lines are too long
|
After Width: | Height: | Size: 12 KiB |
|
|
@ -0,0 +1,16 @@
|
|||
stateDiagram-v2
|
||||
[*] --> starting
|
||||
starting --> idle
|
||||
starting --> error
|
||||
|
||||
idle --> running: userPromptSubmit hook<br/>(user sends prompt)
|
||||
running --> idle: stop hook<br/>(agent finishes)
|
||||
|
||||
running --> error: crash
|
||||
error --> starting: restart
|
||||
|
||||
note right of idle
|
||||
Any state can transition to
|
||||
stopped or killed (not shown)
|
||||
stopped can restart → starting
|
||||
end note
|
||||
File diff suppressed because one or more lines are too long
|
After Width: | Height: | Size: 147 KiB |
Reference in New Issue