update urls

This commit is contained in:
Levi Neely 2026-06-13 11:33:50 +02:00
parent 394b9f3a20
commit 528ea5b3b8
1 changed files with 17 additions and 17 deletions

View File

@ -10,14 +10,14 @@ AnviLLM tackled the first question: can you orchestrate multiple agentic CLIs (C
The next steps build on what we learned here:
- [ollie](https://github.com/lneely/ollie) — a Go library for building LLM agents from scratch, rather than wrapping existing CLIs. Handles backends, tools, sandboxing, and skills. Knows nothing about 9P or orchestration. (Started as an effort to make local Ollama models more useful, and evolved from there.)
- [ollie-9p](https://github.com/lneely/ollie-9p) — the core of the next research phase: how far can we take the idea of filesystem-as-agent-surface? Wraps ollie sessions in 9P, exposing meaningful agent state and interactions as a virtual filesystem. Any program that can read and write files — a frontend, an orchestrator, a supervisor, a shell script — can interact with agents directly. Maybe. Hopefully. We'll see.
- [ollie](https://lneely.de/lkn/ollie) — a Go library for building LLM agents from scratch, rather than wrapping existing CLIs. Handles backends, tools, sandboxing, and skills. Knows nothing about 9P or orchestration. (Started as an effort to make local Ollama models more useful, and evolved from there.)
- [ollie-9p](https://lneely.de/lkn/ollie-9p) — the core of the next research phase: how far can we take the idea of filesystem-as-agent-surface? Wraps ollie sessions in 9P, exposing meaningful agent state and interactions as a virtual filesystem. Any program that can read and write files — a frontend, an orchestrator, a supervisor, a shell script — can interact with agents directly. Maybe. Hopefully. We'll see.
## Architecture
![Architecture](docs/diagrams/architecture.svg?v=2)
**Core:** anvillm (9P daemon), anvilmcp (optional MCP server, see [anvillm-mcp](https://github.com/lneely/anvillm-mcp))
**Core:** anvillm (9P daemon), anvilmcp (optional MCP server, see [anvillm-mcp](https://lneely.de/lkn/anvillm-mcp))
**Frontends:** Any 9P-speaking program — currently Assist (Acme)
**Benefits:** Shared sessions, crash recovery, scriptable via 9P, cross-backend agent communication
@ -27,12 +27,12 @@ The next steps build on what we learned here:
## Requirements
Go 1.21+, [plan9port](https://github.com/lneely/plan9port) (wayland-9pfuse-truncate branch, provides `9pfuse` with truncate fix), tmux, [landrun](https://github.com/zouuup/landrun) (kernel 5.13+), backend ([Claude Code](https://github.com/anthropics/claude-code), [Kiro](https://kiro.dev), or [Ollama](https://ollama.com) + [ollie](https://github.com/lneely/ollie))
Go 1.21+, [plan9port](https://lneely.de/lkn/plan9port) (wayland-9pfuse-truncate branch, provides `9pfuse` with truncate fix), tmux, [landrun](https://github.com/zouuup/landrun) (kernel 5.13+), backend ([Claude Code](https://github.com/anthropics/claude-code), [Kiro](https://kiro.dev), or [Ollama](https://ollama.com) + [ollie](https://lneely.de/lkn/ollie))
## Installation
```sh
git clone https://github.com/lneely/anvillm && cd anvillm && mk
git clone https://lneely.de/lkn/anvillm && cd anvillm && mk
```
**Service integration** (optional, see `services/*/README.md`):
@ -73,7 +73,7 @@ NAMESPACE=/tmp/ns.$USER.:1 Assist
Any program that speaks 9P can be a frontend. The 9P filesystem exposes session management, state, messaging, and configuration as plain files — so building a new frontend is just reading and writing files.
- **Assist** — Acme client ([anvillm-acme](https://github.com/lneely/anvillm-acme))
- **Assist** — Acme client ([anvillm-acme](https://lneely.de/lkn/anvillm-acme))
For web frontends that can't speak 9P directly, `anvilwebgw` bridges HTTP to 9P.
@ -133,7 +133,7 @@ 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))
**Backends:** Claude (`npm install -g @anthropic-ai/claude-code`), Kiro ([kiro.dev](https://kiro.dev)), Ollama (local models via [ollie](https://lneely.de/lkn/ollie))
**Sandbox:** [landrun](https://github.com/zouuup/landrun) (always enabled) — Defaults: CWD/`/tmp`/config (rw), `/usr`/`/lib`/`/bin` (ro+exec), no network
@ -181,9 +181,9 @@ Run local LLMs via Ollama.
ollama pull qwen2.5-coder:7b
```
2. **Ollie CLI:** Install from [github.com/lneely/ollie](https://github.com/lneely/ollie)
2. **Ollie CLI:** Install from [lneely.de/lkn/ollie](https://lneely.de/lkn/ollie)
```sh
git clone https://github.com/lneely/ollie && cd ollie && go install
git clone https://lneely.de/lkn/ollie && cd ollie && go install
```
**Usage:**
@ -230,7 +230,7 @@ Different clients interact with different parts of the filesystem: frontends rea
### Beads
Task tracking is provided by the separate [9beads](https://github.com/lneely/9beads) service, which exposes a `beads/` 9P filesystem.
Task tracking is provided by the separate [9beads](https://lneely.de/lkn/9beads) service, which exposes a `beads/` 9P filesystem.
### Events & Mailbox
@ -365,7 +365,7 @@ The Conductor analyzes dependencies, spawns specialized bots, and delegates work
## MCP Integration
`anvilmcp` ([anvillm-mcp](https://github.com/lneely/anvillm-mcp)) provides sandboxed code execution for MCP clients (Claude Desktop, Kiro, etc.).
`anvilmcp` ([anvillm-mcp](https://lneely.de/lkn/anvillm-mcp)) provides sandboxed code execution for MCP clients (Claude Desktop, Kiro, etc.).
**Note:** anvilmcp is optional. Tools are served by anvillm via 9P and can be invoked directly:
@ -375,7 +375,7 @@ bash <(9p read anvillm/tools/check_inbox.sh)
anvilmcp adds sandbox isolation (landlock/landrun) around execution for MCP clients.
**Install:** See [anvillm-mcp](https://github.com/lneely/anvillm-mcp) for backend-specific setup.
**Install:** See [anvillm-mcp](https://lneely.de/lkn/anvillm-mcp) for backend-specific setup.
See [Code Execution User Guide](docs/code-execution-user-guide.md) for details.
@ -383,11 +383,11 @@ See [Code Execution User Guide](docs/code-execution-user-guide.md) for details.
| Project | Description |
|---------|-------------|
| [9beads](https://github.com/lneely/9beads) | Task management via 9P — beads-based workflow tracking |
| [anvillm-acme](https://github.com/lneely/anvillm-acme) | Acme frontend for session management |
| [9beads-acme](https://github.com/lneely/9beads-acme) | Acme frontend for 9beads task management |
| [anvillm-mcp](https://github.com/lneely/anvillm-mcp) | MCP server with sandboxed execution |
| [agent-skills](https://github.com/lneely/agent-skills) | Discoverable skill definitions for agents |
| [9beads](https://lneely.de/lkn/9beads) | Task management via 9P — beads-based workflow tracking |
| [anvillm-acme](https://lneely.de/lkn/anvillm-acme) | Acme frontend for session management |
| [9beads-acme](https://lneely.de/lkn/9beads-acme) | Acme frontend for 9beads task management |
| [anvillm-mcp](https://lneely.de/lkn/anvillm-mcp) | MCP server with sandboxed execution |
| [agent-skills](https://lneely.de/lkn/agent-skills) | Discoverable skill definitions for agents |
## Troubleshooting