Skip to content

Commit ad5af03

Browse files
committed
Add docs
1 parent 878df0d commit ad5af03

17 files changed

Lines changed: 1237 additions & 1172 deletions

File tree

assets/favicon.png

2.99 KB
Loading

docs/assets/favicon.png

2.99 KB
Loading

docs/assets/logo.png

622 KB
Loading

docs/getting-started/index.md

Lines changed: 19 additions & 22 deletions
Original file line numberDiff line numberDiff line change
@@ -1,52 +1,49 @@
1-
# Getting Started
1+
# 🚀 Getting Started
22

3-
Welcome to **RLM Code** -- the Research Playground and Evaluation OS for Recursive Language Model (RLM) agentic systems. RLM Code provides an interactive TUI-based development environment for building, benchmarking, and optimizing agent workflows through slash commands and natural language.
3+
Welcome to **RLM Code**, the Research Playground and Evaluation OS for Recursive Language Model (RLM) agentic systems. RLM Code provides a unified TUI-based development environment for building, benchmarking, and optimizing agent workflows through slash commands and natural language.
44

55
---
66

7-
## What is RLM Code?
7+
## 🧪 What is RLM Code?
88

99
RLM Code implements the **Recursive Language Model** paradigm from the research paper *"Recursive Language Models"* (Zhang, Kraska, Khattab, 2025). It extends the paper's concepts with:
1010

11-
- **Context-as-variable**: Context is stored as a REPL variable rather than in the token window, enabling unbounded output and token-efficient processing.
12-
- **Deep recursion**: Support for recursion depth > 1, exceeding the paper's original limitation.
13-
- **Multi-paradigm execution**: Pure RLM, CodeAct, and Traditional paradigms with side-by-side comparison.
14-
- **Pluggable observability**: MLflow, OpenTelemetry, LangSmith, LangFuse, and Logfire integrations.
15-
- **Sandbox runtimes**: Local, Docker, Apple Container, Modal, E2B, and Daytona execution environments.
11+
- 🧠 **Context-as-variable**: Context is stored as a REPL variable rather than in the token window, enabling unbounded output and token-efficient processing
12+
- 🔁 **Deep recursion**: Support for recursion depth > 1, exceeding the paper's original limitation
13+
- 🔀 **Multi-paradigm execution**: Pure RLM, CodeAct, and Traditional paradigms with side-by-side comparison
14+
- 📊 **Pluggable observability**: MLflow, OpenTelemetry, LangSmith, LangFuse, and Logfire integrations
15+
- 📦 **Sandbox runtimes**: Local, Docker, Apple Container, Modal, E2B, and Daytona execution environments
1616

1717
---
1818

19-
## Where to Go Next
19+
## 📚 Where to Go Next
2020

2121
| Guide | Description |
2222
|-------|-------------|
23-
| [Installation](installation.md) | System requirements, package installation, optional dependencies, and verification |
24-
| [Quick Start](quickstart.md) | Launch the TUI, connect a model, run your first benchmark, and explore session replay |
25-
| [CLI Reference](cli.md) | Complete reference for both entry points and all 50+ slash commands |
26-
| [Configuration](configuration.md) | Full `rlm_config.yaml` schema, environment variables, and ConfigManager API |
23+
| [📦 Installation](installation.md) | System requirements, package installation, optional dependencies, and verification |
24+
| [Quick Start](quickstart.md) | Launch the TUI, connect a model, run your first benchmark, explore the Research tab |
25+
| [💻 CLI Reference](cli.md) | Complete reference for the entry point and all 50+ slash commands |
26+
| [⚙️ Configuration](configuration.md) | Full `rlm_config.yaml` schema, environment variables, and ConfigManager API |
2727

2828
---
2929

30-
## Quick Overview
30+
## Quick Overview
3131

3232
```bash
3333
# Install
3434
pip install rlm-code
3535

36-
# Launch the standard TUI
36+
# Launch the unified TUI
3737
rlm-code
3838

39-
# Launch the Research TUI directly
40-
rlm-research
41-
4239
# Connect to a model and run a benchmark
4340
/connect anthropic claude-sonnet-4-20250514
4441
/rlm bench preset=dspy_quick
4542
/leaderboard
4643
```
4744

48-
!!! tip "First Time?"
49-
Start with the [Installation](installation.md) guide to set up your environment, then follow the [Quick Start](quickstart.md) for a hands-on walkthrough.
45+
!!! tip "🆕 First Time?"
46+
Start with the [📦 Installation](installation.md) guide to set up your environment, then follow the [Quick Start](quickstart.md) for a hands-on walkthrough.
5047

51-
!!! info "Two TUI Modes"
52-
RLM Code ships with two TUI modes: the **Standard TUI** (multi-pane workspace with chat, files, details, and shell panels) and the **Research TUI** (dark-themed research lab interface with file browser, code viewer, and metrics bar). Use `rlm-code` for the standard mode or `rlm-research` (or `rlm-code --research`) for the research mode.
48+
!!! info "🖥️ Unified TUI"
49+
RLM Code ships a **single TUI** with 5 tabs: **💬 Chat**, **📁 Files**, **📋 Details**, **⚡ Shell**, and **🔬 Research**. Use `rlm-code` to launch, and press `Ctrl+5` to access the Research tab for experiment tracking, benchmarks, and session replay.

docs/getting-started/installation.md

Lines changed: 3 additions & 3 deletions
Original file line numberDiff line numberDiff line change
@@ -1,4 +1,4 @@
1-
# Installation
1+
# 📦 Installation
22

33
This guide covers how to install RLM Code, its optional dependencies, and how to verify your installation.
44

@@ -88,8 +88,8 @@ The multi-pane terminal interface requires Textual:
8888
pip install rlm-code[tui]
8989
```
9090

91-
!!! note "TUI Required for Interactive Mode"
92-
The `textual` package (>= 0.86.0) is required for both the Standard TUI and the Research TUI. Without it, only headless/scripting usage is available.
91+
!!! note "🖥️ TUI Required for Interactive Mode"
92+
The `textual` package (>= 0.86.0) is required for the TUI with all 5 tabs (Chat, Files, Details, Shell, Research). Without it, only headless/scripting usage is available.
9393

9494
### LLM Providers
9595

docs/getting-started/quickstart.md

Lines changed: 40 additions & 38 deletions
Original file line numberDiff line numberDiff line change
@@ -1,18 +1,18 @@
1-
# Quick Start
1+
# Quick Start
22

3-
This guide walks you through launching RLM Code, connecting to an LLM, running your first benchmark, viewing the leaderboard, and exploring session replay -- all in under 10 minutes.
3+
This guide walks you through launching RLM Code, connecting to an LLM, running your first benchmark, viewing the leaderboard, and exploring the Research tab, all in under 10 minutes.
44

55
---
66

7-
## Prerequisites
7+
## Prerequisites
88

99
Before you begin, make sure you have:
1010

11-
- [x] Python 3.10+ installed
12-
- [x] RLM Code installed (`pip install rlm-code[tui,llm-all]`)
13-
- [x] At least one LLM API key (OpenAI, Anthropic, or Gemini) or a local Ollama instance
11+
- [x] 🐍 Python 3.10+ installed
12+
- [x] 📦 RLM Code installed (`pip install rlm-code[tui,llm-all]`)
13+
- [x] 🔑 At least one LLM API key (OpenAI, Anthropic, or Gemini) or a local Ollama instance
1414

15-
!!! tip "Local Models"
15+
!!! tip "🏠 Local Models"
1616
You can use RLM Code entirely with local models via [Ollama](https://ollama.com/). No API keys needed:
1717

1818
```bash
@@ -21,7 +21,7 @@ Before you begin, make sure you have:
2121

2222
---
2323

24-
## Step 1: Launch the TUI
24+
## Step 1: 🚀 Launch the TUI
2525

2626
Navigate to a project directory (not your home directory) and launch:
2727

@@ -30,28 +30,14 @@ mkdir -p ~/projects/rlm-demo && cd ~/projects/rlm-demo
3030
rlm-code
3131
```
3232

33-
!!! warning "Directory Safety Check"
33+
!!! warning "⚠️ Directory Safety Check"
3434
RLM Code performs a safety check on startup. It will warn you if you are running from your home directory, Desktop, Documents, or a system directory. Always run from a dedicated project directory.
3535

36-
You should see the RLM Code TUI with a multi-pane layout: a chat panel, file browser, details panel, and shell.
37-
38-
### Alternative: Research TUI
39-
40-
For a dark-themed research lab interface with file browser, code viewer, and metrics bar:
41-
42-
```bash
43-
rlm-research
44-
```
45-
46-
Or use the flag:
47-
48-
```bash
49-
rlm-code --research
50-
```
36+
You should see the **RLM Research Lab** TUI with 5 tabs: 💬 Chat, 📁 Files, 📋 Details, ⚡ Shell, and 🔬 Research. The Chat tab is active by default.
5137

5238
---
5339

54-
## Step 2: Initialize Your Project
40+
## Step 2: 📁 Initialize Your Project
5541

5642
Initialize a project configuration file:
5743

@@ -63,7 +49,7 @@ This creates an `rlm_config.yaml` in your current directory with default setting
6349

6450
---
6551

66-
## Step 3: Connect to a Model
52+
## Step 3: 🔗 Connect to a Model
6753

6854
Use the `/connect` command to connect to an LLM provider:
6955

@@ -125,7 +111,7 @@ This shows the current model, provider, connection status, sandbox runtime, and
125111

126112
---
127113

128-
## Step 4: Run a Benchmark
114+
## Step 4: 🏆 Run a Benchmark
129115

130116
RLM Code ships with 10+ built-in benchmark presets. Start with the quick DSPy smoke test:
131117

@@ -178,7 +164,7 @@ Supported formats include explicit preset mappings, Pydantic-style dataset cases
178164

179165
---
180166

181-
## Step 5: View the Leaderboard
167+
## Step 5: 📊 View the Leaderboard
182168

183169
After running benchmarks, view aggregated results on the leaderboard:
184170

@@ -209,7 +195,7 @@ Available ranking metrics: `reward`, `completion_rate`, `steps`, `tokens`, `cost
209195

210196
---
211197

212-
## Step 6: Compare Paradigms
198+
## Step 6: 🔀 Compare Paradigms
213199

214200
Run the same task through multiple paradigms and compare:
215201

@@ -227,7 +213,7 @@ Use the comparison command for direct A/B analysis:
227213

228214
---
229215

230-
## Step 7: Session Replay
216+
## Step 7: Session Replay
231217

232218
Every RLM run generates a trajectory that can be replayed step by step.
233219

@@ -255,7 +241,7 @@ Session replay supports both JSONL trajectory files and JSON snapshot files.
255241

256242
---
257243

258-
## Step 8: Explore Slash Commands
244+
## Step 8: ⌨️ Explore Slash Commands
259245

260246
RLM Code has 50+ slash commands. Here are the most useful ones to explore next:
261247

@@ -310,7 +296,22 @@ RLM Code has 50+ slash commands. Here are the most useful ones to explore next:
310296

311297
---
312298

313-
## Full Workflow Example
299+
## 🔬 Step 9: Explore the Research Tab
300+
301+
After running a benchmark, press `Ctrl+5` to switch to the **🔬 Research** tab:
302+
303+
- **📊 Dashboard**: See run metrics, reward sparkline, and summary
304+
- **📈 Trajectory**: Step-by-step breakdown of agent actions and rewards
305+
- **🏆 Benchmarks**: Leaderboard table from all your runs
306+
- **⏪ Replay**: Step-through controls for time-travel debugging
307+
- **📡 Events**: Live event stream from the RLM event bus
308+
309+
!!! tip "🔬 Research Tab"
310+
The Research tab updates automatically when you run `/rlm bench` or `/rlm run` commands. No manual refresh needed!
311+
312+
---
313+
314+
## 🎯 Full Workflow Example
314315

315316
Here is a complete workflow from start to finish:
316317

@@ -363,10 +364,11 @@ rlm-code
363364

364365
---
365366

366-
## What's Next?
367+
## 📚 What's Next?
367368

368-
- **[CLI Reference](cli.md)**: Complete documentation for all commands and flags
369-
- **[Configuration](configuration.md)**: Customize every aspect of RLM Code via `rlm_config.yaml`
370-
- Explore the [Core Engine](../core/index.md) documentation for the RLM Runner, Environments, and Event System
371-
- Set up [Observability](../observability/index.md) with MLflow, OpenTelemetry, or your preferred platform
372-
- Learn about [Sandbox Runtimes](../sandbox/index.md) for isolated code execution with Docker, Modal, or E2B
369+
- 💻 **[CLI Reference](cli.md)**: Complete documentation for all commands and flags
370+
- ⚙️ **[Configuration](configuration.md)**: Customize every aspect of RLM Code via `rlm_config.yaml`
371+
- 🧠 **[Core Engine](../core/index.md)**: RLM Runner, Environments, and Event System
372+
- 🔬 **[Research Tab](../tui/research.md)**: Deep dive into the experiment tracking interface
373+
- 📊 **[Observability](../observability/index.md)**: MLflow, OpenTelemetry, LangSmith, LangFuse, Logfire
374+
- 📦 **[Sandbox Runtimes](../sandbox/index.md)**: Docker, Modal, E2B for isolated code execution

0 commit comments

Comments
 (0)