code-graph-llm 2.1.0 → 2.1.2
This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
- package/README.md +50 -6
- package/package.json +1 -1
package/README.md
CHANGED
|
@@ -1,12 +1,12 @@
|
|
|
1
|
-
# CODE-GRAPH (v2.
|
|
1
|
+
# CODE-GRAPH (v2.1.2)
|
|
2
2
|
|
|
3
3
|
A language-agnostic, ultra-compact codebase mapper and **agent memory system** designed specifically for LLM agents. It optimizes context and token usage while enabling agents to learn from their own mistakes across sessions.
|
|
4
4
|
|
|
5
|
-
## 🚀 New in v2.
|
|
6
|
-
- **
|
|
7
|
-
- **
|
|
8
|
-
- **
|
|
9
|
-
- **Production
|
|
5
|
+
## 🚀 New in v2.1: Intelligent Reflection & Hard-Soft Enforcement
|
|
6
|
+
- **Project Initializer:** Automated bootstrapping with `code-graph init`.
|
|
7
|
+
- **Intelligent Reflection:** Deduplication and categorization in `PROJECT_REFLECTIONS.md`.
|
|
8
|
+
- **Commit Advisories:** Soft-nudge git hooks to remind agents of missing reflections.
|
|
9
|
+
- **Production-Ready Core:** Refactored Service-based architecture with full async support.
|
|
10
10
|
|
|
11
11
|
## Features
|
|
12
12
|
- **Structural Knowledge Graph:** Captures `imports`, `requires`, `extends`, and `implements`.
|
|
@@ -68,6 +68,50 @@ The `llm-code-graph.md` file provides a high-level map and structural graph for
|
|
|
68
68
|
### 3. Example System Prompt
|
|
69
69
|
> "Before acting, read `llm-code-graph.md`. Follow the protocol in `AGENT_RULES.md`. If you encounter a bug or an environment quirk, use the `code-graph reflect` tool to record the lesson in `PROJECT_REFLECTIONS.md`."
|
|
70
70
|
|
|
71
|
+
## 🤖 Agent-Specific Integration
|
|
72
|
+
|
|
73
|
+
Maximize efficiency by pointing your agent directly to the `llm-code-graph.md` and `AGENT_RULES.md` files.
|
|
74
|
+
|
|
75
|
+
### Roo Code (Cline)
|
|
76
|
+
Add this to your `.clinerules` file:
|
|
77
|
+
```markdown
|
|
78
|
+
Before starting any task:
|
|
79
|
+
1. Read `llm-code-graph.md` for project structure.
|
|
80
|
+
2. Read `AGENT_RULES.md` for operational protocol.
|
|
81
|
+
3. Read `PROJECT_REFLECTIONS.md` for past lessons.
|
|
82
|
+
```
|
|
83
|
+
|
|
84
|
+
### Cursor / Windsurf
|
|
85
|
+
Add to `.cursorrules` or `.windsurfrules`:
|
|
86
|
+
```text
|
|
87
|
+
Always reference `llm-code-graph.md` to understand the codebase architecture before performing edits. Strictly follow the "Reflection Cycle" in `AGENT_RULES.md`.
|
|
88
|
+
```
|
|
89
|
+
|
|
90
|
+
### GitHub Copilot
|
|
91
|
+
Add to `.github/copilot-instructions.md`:
|
|
92
|
+
```markdown
|
|
93
|
+
When answering questions about this codebase, prioritize the mapping data in `llm-code-graph.md`. If a task fails or a mistake is corrected, suggest a reflection entry for `PROJECT_REFLECTIONS.md`.
|
|
94
|
+
```
|
|
95
|
+
|
|
96
|
+
### Gemini CLI
|
|
97
|
+
Create a `GEMINI.md` file (if not already using `AGENT_RULES.md`):
|
|
98
|
+
```markdown
|
|
99
|
+
- Always read `llm-code-graph.md` as the primary source of truth.
|
|
100
|
+
- Use the `run_shell_command` tool to execute `code-graph reflect` after fixing non-obvious bugs.
|
|
101
|
+
```
|
|
102
|
+
|
|
103
|
+
### OpenCode / OpenDevin / Aider
|
|
104
|
+
In your project instructions or session startup:
|
|
105
|
+
> "Read `llm-code-graph.md` for a high-level overview. If you discover a project-specific quirk, record it using `code-graph reflect <CAT> <LESSON>`."
|
|
106
|
+
|
|
107
|
+
### Codex / GPT-Engineer / AutoGPT
|
|
108
|
+
Add to your project prompt or `prompts.md`:
|
|
109
|
+
> "Use `llm-code-graph.md` to navigate the file structure. Strictly adhere to the reflection cycle in `AGENT_RULES.md` to ensure project memory is maintained across iterations."
|
|
110
|
+
|
|
111
|
+
### Generic Agent (System Prompt)
|
|
112
|
+
For any other agent, add this to your system instructions:
|
|
113
|
+
> "This project uses `code-graph-llm` for context management. Always consult `llm-code-graph.md` before exploring files. If you learn something new about the environment (e.g., shell quirks, library oddities), persist that knowledge using the `code-graph reflect` tool."
|
|
114
|
+
|
|
71
115
|
## How it works
|
|
72
116
|
1. **File Scanning:** Recursively walks the directory, ignoring patterns in `.gitignore`.
|
|
73
117
|
2. **Context Extraction:** Scans for classes, functions, and variables while ignoring matches in comments.
|