tenseleyflow/loader / dbbb638

Browse files

update docs, tracking

Authored by mfwolffe <wolffemf@dukes.jmu.edu>
SHA
dbbb638302a575e5c0ab3c0deba296d4983cba1d
Parents
f91124a
Tree
3ea744b

2 changed files

StatusFile+-
M .gitignore 1 0
M README.md 25 60
.gitignoremodified
@@ -53,3 +53,4 @@ Thumbs.db
5353
5454
 # Local docs
5555
 docs/
56
+.fackr/
README.mdmodified
@@ -1,80 +1,45 @@
11
 # Loader
22
3
-A local-first, open-source agentic coding assistant. Think Claude Code, but running on your own hardware with open-weight LLMs.
3
+Local agentic coding assistant. Runs on your hardware with local LLMs.
44
5
-## Vision
5
+## Install
66
7
-Loader provides a scaffold for running agentic AI coding assistants with:
8
-- **Local LLM backends** - Ollama, llama.cpp, vLLM, or any OpenAI-compatible API
9
-- **Tool system** - File operations, shell commands, code search, and more
10
-- **Agent loop** - Think → Plan → Act → Observe → Repeat
11
-- **Streaming CLI** - Real-time responses in your terminal
12
-
13
-## Architecture
14
-
15
-```
16
-┌─────────────────────────────────────────────────────────┐
17
-│                         CLI                              │
18
-│  (Rich terminal UI, streaming, conversation history)    │
19
-└─────────────────────────────────────────────────────────┘
20
-                            │
21
-┌─────────────────────────────────────────────────────────┐
22
-│                      Agent Loop                          │
23
-│  (Think → Select Tool → Execute → Observe → Repeat)     │
24
-└─────────────────────────────────────────────────────────┘
25
-                            │
26
-        ┌───────────────────┼───────────────────┐
27
-        ▼                   ▼                   ▼
28
-┌───────────────┐   ┌───────────────┐   ┌───────────────┐
29
-│    Tools      │   │  LLM Backend  │   │    Context    │
30
-│  - Read       │   │  - Ollama     │   │  - Messages   │
31
-│  - Write      │   │  - llama.cpp  │   │  - Files      │
32
-│  - Bash       │   │  - vLLM       │   │  - Codebase   │
33
-│  - Grep       │   │  - OpenAI API │   │  - Memory     │
34
-│  - Glob       │   └───────────────┘   └───────────────┘
35
-└───────────────┘
7
+```bash
8
+pip install -e .
369
 ```
3710
38
-## Features (Planned)
39
-
40
-- [ ] Core agent loop with tool use
41
-- [ ] File operations (read, write, edit)
42
-- [ ] Shell command execution
43
-- [ ] Code search (grep, glob, ripgrep)
44
-- [ ] Ollama backend integration
45
-- [ ] Streaming responses
46
-- [ ] Conversation history
47
-- [ ] Context window management
48
-- [ ] Multi-file awareness
49
-- [ ] Git integration
50
-
5111
 ## Requirements
5212
5313
 - Python 3.11+
54
-- Ollama (or other LLM backend)
55
-- A capable GPU (recommended: 8GB+ VRAM for 7B models)
14
+- Ollama running (`ollama serve`)
15
+- A model pulled (`ollama pull llama3.1:8b`)
5616
57
-## Quick Start
17
+## Usage
5818
5919
 ```bash
60
-# Install
61
-pip install -e .
20
+# Interactive mode
21
+loader
6222
63
-# Run with Ollama backend
64
-loader --backend ollama --model llama3.1:8b
23
+# Single prompt
24
+loader "Read main.py and explain it"
6525
66
-# Or with a specific task
67
-loader "Help me refactor this function"
26
+# Skip confirmation prompts
27
+loader -y "Create a hello.py file"
28
+
29
+# Use different model
30
+loader -m qwen2.5:7b
6831
 ```
6932
70
-## Supported Models
33
+**In interactive mode:** type prompts, `clear` to reset, `exit` to quit.
34
+
35
+## Tools
7136
72
-Any model that can do tool/function calling:
73
-- Llama 3.1 (8B, 70B) - Excellent tool use
74
-- Qwen 2.5 (7B, 14B, 32B) - Great coding ability
75
-- Mistral/Mixtral - Good general purpose
76
-- DeepSeek Coder - Specialized for code
77
-- CodeLlama - Code-focused
37
+- `read` - read files
38
+- `write` - write files
39
+- `edit` - find/replace in files
40
+- `glob` - find files by pattern
41
+- `grep` - search file contents
42
+- `bash` - run shell commands
7843
7944
 ## License
8045