claude-self-reflect 2.8.0 โ 2.8.1
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/.claude/agents/docker-orchestrator.md +1 -1
- package/.claude/agents/import-debugger.md +1 -1
- package/.claude/agents/mcp-integration.md +1 -1
- package/.claude/agents/qdrant-specialist.md +1 -1
- package/.claude/agents/search-optimizer.md +1 -1
- package/README.md +103 -146
- package/mcp-server/src/project_resolver.py +2 -2
- package/mcp-server/src/server.py +46 -34
- package/package.json +1 -1
- package/scripts/import-latest.py +124 -0
- package/scripts/import-old-format.py +171 -0
|
@@ -4,7 +4,7 @@ description: Docker Compose orchestration expert for container management, servi
|
|
|
4
4
|
tools: Read, Edit, Bash, Grep, LS
|
|
5
5
|
---
|
|
6
6
|
|
|
7
|
-
You are a Docker orchestration specialist for the
|
|
7
|
+
You are a Docker orchestration specialist for the claude-self-reflect project. You manage multi-container deployments, monitor service health, and troubleshoot container issues.
|
|
8
8
|
|
|
9
9
|
## Project Context
|
|
10
10
|
- Main stack: Qdrant vector database + MCP server + Python importer
|
|
@@ -4,7 +4,7 @@ description: Import pipeline debugging specialist for JSONL processing, Python s
|
|
|
4
4
|
tools: Read, Edit, Bash, Grep, Glob, LS
|
|
5
5
|
---
|
|
6
6
|
|
|
7
|
-
You are an import pipeline debugging expert for the
|
|
7
|
+
You are an import pipeline debugging expert for the claude-self-reflect project. You specialize in troubleshooting JSONL file processing, Python import scripts, and conversation chunking strategies.
|
|
8
8
|
|
|
9
9
|
## Project Context
|
|
10
10
|
- Processes Claude Desktop logs from ~/.claude/projects/
|
|
@@ -4,7 +4,7 @@ description: MCP (Model Context Protocol) server development expert for Claude D
|
|
|
4
4
|
tools: Read, Edit, Bash, Grep, Glob, WebFetch
|
|
5
5
|
---
|
|
6
6
|
|
|
7
|
-
You are an MCP server development specialist for the
|
|
7
|
+
You are an MCP server development specialist for the claude-self-reflect project. You handle Claude Desktop integration, implement MCP tools, and ensure seamless communication between Claude and the vector database.
|
|
8
8
|
|
|
9
9
|
## Project Context
|
|
10
10
|
- MCP server: claude-self-reflection
|
|
@@ -4,7 +4,7 @@ description: Qdrant vector database expert for collection management, troublesho
|
|
|
4
4
|
tools: Read, Bash, Grep, Glob, LS, WebFetch
|
|
5
5
|
---
|
|
6
6
|
|
|
7
|
-
You are a Qdrant vector database specialist for the
|
|
7
|
+
You are a Qdrant vector database specialist for the claude-self-reflect project. Your expertise covers collection management, vector search optimization, and embedding strategies.
|
|
8
8
|
|
|
9
9
|
## Project Context
|
|
10
10
|
- The system uses Qdrant for storing conversation embeddings from Claude Desktop logs
|
|
@@ -4,7 +4,7 @@ description: Search quality optimization expert for improving semantic search ac
|
|
|
4
4
|
tools: Read, Edit, Bash, Grep, Glob, WebFetch
|
|
5
5
|
---
|
|
6
6
|
|
|
7
|
-
You are a search optimization specialist for the
|
|
7
|
+
You are a search optimization specialist for the claude-self-reflect project. You improve semantic search quality, tune parameters, and analyze embedding model performance.
|
|
8
8
|
|
|
9
9
|
## Project Context
|
|
10
10
|
- Current baseline: 66.1% search accuracy with Voyage AI
|
package/README.md
CHANGED
|
@@ -24,67 +24,12 @@
|
|
|
24
24
|
|
|
25
25
|
Give Claude perfect memory of all your conversations. Search past discussions instantly. Never lose context again.
|
|
26
26
|
|
|
27
|
-
|
|
27
|
+
**๐ 100% Local by Default** โข **โก Blazing Fast Search** โข **๐ Zero Configuration** โข **๐ญ Production Ready**
|
|
28
28
|
|
|
29
|
-
|
|
29
|
+
## ๐ Quick Install
|
|
30
30
|
|
|
31
|
-
**Zero Configuration** - Works immediately after installation. Smart auto-detection handles everything. No manual setup, no environment variables, just install and use.
|
|
32
|
-
|
|
33
|
-
**Production Ready** - Battle-tested with 600+ conversations across 24 projects. Handles mixed embedding types automatically. Scales from personal use to team deployments.
|
|
34
|
-
|
|
35
|
-
## Table of Contents
|
|
36
|
-
|
|
37
|
-
- [What You Get](#what-you-get)
|
|
38
|
-
- [Requirements](#requirements)
|
|
39
|
-
- [Quick Install/Uninstall](#quick-installuninstall)
|
|
40
|
-
- [The Magic](#the-magic)
|
|
41
|
-
- [Before & After](#before--after)
|
|
42
|
-
- [Real Examples](#real-examples-that-made-us-build-this)
|
|
43
|
-
- [How It Works](#how-it-works)
|
|
44
|
-
- [Import Architecture](#import-architecture)
|
|
45
|
-
- [Using It](#using-it)
|
|
46
|
-
- [Key Features](#key-features)
|
|
47
|
-
- [Performance](#performance)
|
|
48
|
-
- [Configuration](#configuration)
|
|
49
|
-
- [Technical Stack](#the-technical-stack)
|
|
50
|
-
- [Problems](#problems)
|
|
51
|
-
- [What's New](#whats-new)
|
|
52
|
-
- [Advanced Topics](#advanced-topics)
|
|
53
|
-
- [Contributors](#contributors)
|
|
54
|
-
|
|
55
|
-
## What You Get
|
|
56
|
-
|
|
57
|
-
Ask Claude about past conversations. Get actual answers. **100% local by default** - your conversations never leave your machine. Cloud-enhanced search available when you need it.
|
|
58
|
-
|
|
59
|
-
**Proven at Scale**: Successfully indexed 682 conversation files with 100% reliability. No data loss, no corruption, just seamless conversation memory that works.
|
|
60
|
-
|
|
61
|
-
**Before**: "I don't have access to previous conversations"
|
|
62
|
-
**After**:
|
|
63
|
-
```
|
|
64
|
-
reflection-specialist(Search FastEmbed vs cloud embedding decision)
|
|
65
|
-
โฟ Done (3 tool uses ยท 8.2k tokens ยท 12.4s)
|
|
66
|
-
|
|
67
|
-
"Found it! Yesterday we decided on FastEmbed for local mode - better privacy,
|
|
68
|
-
no API calls, 384-dimensional embeddings. Works offline too."
|
|
69
|
-
```
|
|
70
|
-
|
|
71
|
-
The reflection specialist is a specialized sub-agent that Claude automatically spawns when you ask about past conversations. It searches your conversation history in its own isolated context, keeping your main chat clean and focused.
|
|
72
|
-
|
|
73
|
-
Your conversations become searchable. Your decisions stay remembered. Your context persists.
|
|
74
|
-
|
|
75
|
-
## Requirements
|
|
76
|
-
|
|
77
|
-
- **Docker Desktop** (macOS/Windows) or **Docker Engine** (Linux)
|
|
78
|
-
- **Node.js** 16+ (for the setup wizard)
|
|
79
|
-
- **Claude Desktop** app
|
|
80
|
-
|
|
81
|
-
## Quick Install/Uninstall
|
|
82
|
-
|
|
83
|
-
### Install
|
|
84
|
-
|
|
85
|
-
#### Local Mode (Default - Your Data Stays Private)
|
|
86
31
|
```bash
|
|
87
|
-
# Install and run automatic setup
|
|
32
|
+
# Install and run automatic setup (5 minutes, everything automatic)
|
|
88
33
|
npm install -g claude-self-reflect
|
|
89
34
|
claude-self-reflect setup
|
|
90
35
|
|
|
@@ -93,11 +38,12 @@ claude-self-reflect setup
|
|
|
93
38
|
# โ
Configure everything automatically
|
|
94
39
|
# โ
Install the MCP in Claude Code
|
|
95
40
|
# โ
Start monitoring for new conversations
|
|
96
|
-
# โ
Verify the reflection tools work
|
|
97
41
|
# ๐ Keep all data local - no API keys needed
|
|
98
42
|
```
|
|
99
43
|
|
|
100
|
-
|
|
44
|
+
<details open>
|
|
45
|
+
<summary>๐ก Cloud Mode (Better Search Accuracy)</summary>
|
|
46
|
+
|
|
101
47
|
```bash
|
|
102
48
|
# Step 1: Get your free Voyage AI key
|
|
103
49
|
# Sign up at https://www.voyageai.com/ - it takes 30 seconds
|
|
@@ -108,17 +54,17 @@ claude-self-reflect setup --voyage-key=YOUR_ACTUAL_KEY_HERE
|
|
|
108
54
|
```
|
|
109
55
|
*Note: Cloud mode provides more accurate semantic search but sends conversation data to Voyage AI for processing.*
|
|
110
56
|
|
|
111
|
-
|
|
57
|
+
</details>
|
|
112
58
|
|
|
113
|
-
## The Magic
|
|
59
|
+
## โจ The Magic
|
|
114
60
|
|
|
115
61
|
|
|
116
62
|
|
|
117
|
-
## Before & After
|
|
63
|
+
## ๐ Before & After
|
|
118
64
|
|
|
119
65
|
|
|
120
66
|
|
|
121
|
-
## Real Examples
|
|
67
|
+
## ๐ฌ Real Examples
|
|
122
68
|
|
|
123
69
|
```
|
|
124
70
|
You: "What was that PostgreSQL optimization we figured out?"
|
|
@@ -137,41 +83,7 @@ Claude: "3 conversations found:
|
|
|
137
83
|
- Nov 20: Added rate limiting per authenticated connection"
|
|
138
84
|
```
|
|
139
85
|
|
|
140
|
-
##
|
|
141
|
-
|
|
142
|
-
Your conversations โ Vector embeddings โ Semantic search โ Claude remembers
|
|
143
|
-
|
|
144
|
-
Technical details exist. You don't need them to start.
|
|
145
|
-
|
|
146
|
-
## Import Architecture
|
|
147
|
-
|
|
148
|
-
Here's how your conversations get imported and prioritized:
|
|
149
|
-
|
|
150
|
-
|
|
151
|
-
|
|
152
|
-
**The system intelligently processes your conversations:**
|
|
153
|
-
- Runs every 60 seconds checking for new conversations
|
|
154
|
-
- Processes newest conversations first (delta import pattern)
|
|
155
|
-
- Maintains low memory usage (<50MB) through streaming
|
|
156
|
-
- Handles up to 5 files per cycle to prevent blocking
|
|
157
|
-
|
|
158
|
-
**HOT/WARM/COLD Intelligent Prioritization:**
|
|
159
|
-
- **๐ฅ HOT** (< 5 minutes): Switches to 2-second intervals for near real-time import
|
|
160
|
-
- **๐ก๏ธ WARM** (< 24 hours): Normal priority with starvation prevention (urgent after 30 min wait)
|
|
161
|
-
- **โ๏ธ COLD** (> 24 hours): Batch processed, max 5 per cycle to prevent blocking new content
|
|
162
|
-
- Files are categorized by age and processed with priority queuing to ensure newest content gets imported quickly while preventing older files from being starved
|
|
163
|
-
|
|
164
|
-
## Using It
|
|
165
|
-
|
|
166
|
-
Once installed, just talk naturally:
|
|
167
|
-
|
|
168
|
-
- "What did we discuss about database optimization?"
|
|
169
|
-
- "Find our debugging session from last week"
|
|
170
|
-
- "Remember this solution for next time"
|
|
171
|
-
|
|
172
|
-
The reflection specialist automatically activates. No special commands needed.
|
|
173
|
-
|
|
174
|
-
## Key Features
|
|
86
|
+
## ๐ฏ Key Features
|
|
175
87
|
|
|
176
88
|
### Project-Scoped Search
|
|
177
89
|
Searches are **project-aware by default**. Claude automatically searches within your current project:
|
|
@@ -189,16 +101,37 @@ Claude: [Searches across ALL your projects]
|
|
|
189
101
|
### โฑ๏ธ Memory Decay
|
|
190
102
|
Recent conversations matter more. Old ones fade. Like your brain, but reliable.
|
|
191
103
|
|
|
192
|
-
###
|
|
193
|
-
- **Search**: <3ms average response time
|
|
194
|
-
- **
|
|
195
|
-
- **
|
|
196
|
-
- **
|
|
197
|
-
|
|
198
|
-
|
|
104
|
+
### โก Performance at Scale
|
|
105
|
+
- **Search**: <3ms average response time
|
|
106
|
+
- **Scale**: 600+ conversations across 24 projects
|
|
107
|
+
- **Reliability**: 100% indexing success rate
|
|
108
|
+
- **Memory**: 96% reduction from v2.5.15
|
|
109
|
+
|
|
110
|
+
## ๐๏ธ Architecture
|
|
111
|
+
|
|
112
|
+
|
|
113
|
+
|
|
114
|
+
<details>
|
|
115
|
+
<summary>๐ฅ HOT/WARM/COLD Intelligent Prioritization</summary>
|
|
116
|
+
|
|
117
|
+
- **๐ฅ HOT** (< 5 minutes): 2-second intervals for near real-time import
|
|
118
|
+
- **๐ก๏ธ WARM** (< 24 hours): Normal priority with starvation prevention
|
|
119
|
+
- **โ๏ธ COLD** (> 24 hours): Batch processed to prevent blocking
|
|
120
|
+
|
|
121
|
+
Files are categorized by age and processed with priority queuing to ensure newest content gets imported quickly while preventing older files from being starved.
|
|
122
|
+
|
|
123
|
+
</details>
|
|
124
|
+
|
|
125
|
+
## ๐ ๏ธ Requirements
|
|
126
|
+
|
|
127
|
+
- **Docker Desktop** (macOS/Windows) or **Docker Engine** (Linux)
|
|
128
|
+
- **Node.js** 16+ (for the setup wizard)
|
|
129
|
+
- **Claude Desktop** app
|
|
199
130
|
|
|
131
|
+
## ๐ Documentation
|
|
200
132
|
|
|
201
|
-
|
|
133
|
+
<details>
|
|
134
|
+
<summary>๐ง Technical Stack</summary>
|
|
202
135
|
|
|
203
136
|
- **Vector DB**: Qdrant (local, your data stays yours)
|
|
204
137
|
- **Embeddings**:
|
|
@@ -207,18 +140,62 @@ Recent conversations matter more. Old ones fade. Like your brain, but reliable.
|
|
|
207
140
|
- **MCP Server**: Python + FastMCP
|
|
208
141
|
- **Search**: Semantic similarity with time decay
|
|
209
142
|
|
|
210
|
-
|
|
143
|
+
</details>
|
|
144
|
+
|
|
145
|
+
<details>
|
|
146
|
+
<summary>๐ Advanced Topics</summary>
|
|
147
|
+
|
|
148
|
+
- [Performance tuning](docs/performance-guide.md)
|
|
149
|
+
- [Security & privacy](docs/security.md)
|
|
150
|
+
- [Windows setup](docs/windows-setup.md)
|
|
151
|
+
- [Architecture details](docs/architecture-details.md)
|
|
152
|
+
- [Contributing](CONTRIBUTING.md)
|
|
153
|
+
|
|
154
|
+
</details>
|
|
155
|
+
|
|
156
|
+
<details>
|
|
157
|
+
<summary>๐ Troubleshooting</summary>
|
|
211
158
|
|
|
212
159
|
- [Troubleshooting Guide](docs/troubleshooting.md)
|
|
213
160
|
- [GitHub Issues](https://github.com/ramakay/claude-self-reflect/issues)
|
|
214
161
|
- [Discussions](https://github.com/ramakay/claude-self-reflect/discussions)
|
|
215
162
|
|
|
216
|
-
|
|
163
|
+
</details>
|
|
164
|
+
|
|
165
|
+
<details>
|
|
166
|
+
<summary>๐๏ธ Uninstall</summary>
|
|
167
|
+
|
|
168
|
+
For complete uninstall instructions, see [docs/UNINSTALL.md](docs/UNINSTALL.md).
|
|
169
|
+
|
|
170
|
+
Quick uninstall:
|
|
171
|
+
```bash
|
|
172
|
+
# Remove MCP server
|
|
173
|
+
claude mcp remove claude-self-reflect
|
|
174
|
+
|
|
175
|
+
# Stop Docker containers
|
|
176
|
+
docker-compose down
|
|
177
|
+
|
|
178
|
+
# Uninstall npm package
|
|
179
|
+
npm uninstall -g claude-self-reflect
|
|
180
|
+
```
|
|
181
|
+
|
|
182
|
+
</details>
|
|
183
|
+
|
|
184
|
+
## ๐ฆ What's New
|
|
185
|
+
|
|
186
|
+
<details>
|
|
187
|
+
<summary>๐ v2.8.0 - Latest Release</summary>
|
|
217
188
|
|
|
218
|
-
|
|
219
|
-
|
|
189
|
+
- **๐ง Fixed MCP Indexing**: Now correctly shows 97.1% progress (was showing 0%)
|
|
190
|
+
- **๐ฅ HOT/WARM/COLD**: Intelligent file prioritization for near real-time imports
|
|
191
|
+
- **๐ Enhanced Monitoring**: Real-time status with visual indicators
|
|
220
192
|
|
|
221
|
-
|
|
193
|
+
</details>
|
|
194
|
+
|
|
195
|
+
<details>
|
|
196
|
+
<summary>โจ v2.5.19 - Metadata Enrichment</summary>
|
|
197
|
+
|
|
198
|
+
### For Existing Users
|
|
222
199
|
```bash
|
|
223
200
|
# Update to latest version
|
|
224
201
|
npm update -g claude-self-reflect
|
|
@@ -231,50 +208,30 @@ claude-self-reflect setup
|
|
|
231
208
|
docker compose run --rm importer python /app/scripts/delta-metadata-update-safe.py
|
|
232
209
|
```
|
|
233
210
|
|
|
234
|
-
|
|
211
|
+
### What You Get
|
|
235
212
|
- `search_by_concept("docker")` - Find conversations by topic
|
|
236
213
|
- `search_by_file("server.py")` - Find conversations that touched specific files
|
|
237
214
|
- Better search accuracy with metadata-based filtering
|
|
238
215
|
|
|
239
|
-
|
|
216
|
+
</details>
|
|
217
|
+
|
|
218
|
+
<details>
|
|
219
|
+
<summary>๐ Release History</summary>
|
|
240
220
|
|
|
241
|
-
- **v2.5.19** - Metadata Enrichment! Search by concepts, files, and tools. [Full release notes](docs/releases/v2.5.19-RELEASE-NOTES.md)
|
|
242
221
|
- **v2.5.18** - Security dependency updates
|
|
243
|
-
- **v2.5.17** - Critical CPU fix and memory limit adjustment
|
|
244
|
-
- **v2.5.16** -
|
|
222
|
+
- **v2.5.17** - Critical CPU fix and memory limit adjustment
|
|
223
|
+
- **v2.5.16** - Initial streaming importer with CPU throttling
|
|
245
224
|
- **v2.5.15** - Critical bug fixes and collection creation improvements
|
|
246
|
-
- **v2.5.14** - Async importer collection fix
|
|
247
|
-
- **v2.5.11** - Critical cloud mode fix
|
|
248
|
-
- **v2.5.10** - Emergency hotfix for MCP server startup
|
|
249
|
-
- **v2.5.6** - Tool Output Extraction
|
|
225
|
+
- **v2.5.14** - Async importer collection fix
|
|
226
|
+
- **v2.5.11** - Critical cloud mode fix
|
|
227
|
+
- **v2.5.10** - Emergency hotfix for MCP server startup
|
|
228
|
+
- **v2.5.6** - Tool Output Extraction
|
|
250
229
|
|
|
251
230
|
[Full changelog](docs/release-history.md)
|
|
252
231
|
|
|
253
|
-
|
|
254
|
-
|
|
255
|
-
- [Performance tuning](docs/performance-guide.md)
|
|
256
|
-
- [Security & privacy](docs/security.md)
|
|
257
|
-
- [Windows setup](docs/windows-setup.md)
|
|
258
|
-
- [Architecture details](docs/architecture-details.md)
|
|
259
|
-
- [Contributing](CONTRIBUTING.md)
|
|
260
|
-
|
|
261
|
-
### Uninstall
|
|
262
|
-
|
|
263
|
-
For complete uninstall instructions, see [docs/UNINSTALL.md](docs/UNINSTALL.md).
|
|
264
|
-
|
|
265
|
-
Quick uninstall:
|
|
266
|
-
```bash
|
|
267
|
-
# Remove MCP server
|
|
268
|
-
claude mcp remove claude-self-reflect
|
|
269
|
-
|
|
270
|
-
# Stop Docker containers
|
|
271
|
-
docker-compose down
|
|
272
|
-
|
|
273
|
-
# Uninstall npm package
|
|
274
|
-
npm uninstall -g claude-self-reflect
|
|
275
|
-
```
|
|
232
|
+
</details>
|
|
276
233
|
|
|
277
|
-
## Contributors
|
|
234
|
+
## ๐ฅ Contributors
|
|
278
235
|
|
|
279
236
|
Special thanks to our contributors:
|
|
280
237
|
- **[@TheGordon](https://github.com/TheGordon)** - Fixed timestamp parsing (#10)
|
|
@@ -283,4 +240,4 @@ Special thanks to our contributors:
|
|
|
283
240
|
|
|
284
241
|
---
|
|
285
242
|
|
|
286
|
-
Built with โค๏ธ
|
|
243
|
+
Built with โค๏ธ by [ramakay](https://github.com/ramakay) for the Claude community.
|
|
@@ -54,7 +54,7 @@ class ProjectResolver:
|
|
|
54
54
|
4. Fuzzy matching on collection names
|
|
55
55
|
|
|
56
56
|
Args:
|
|
57
|
-
user_project_name: User-provided project name (e.g., "
|
|
57
|
+
user_project_name: User-provided project name (e.g., "example-project", "Example-Project", full path)
|
|
58
58
|
|
|
59
59
|
Returns:
|
|
60
60
|
List of collection names that match the project
|
|
@@ -362,7 +362,7 @@ class ProjectResolver:
|
|
|
362
362
|
|
|
363
363
|
Examples:
|
|
364
364
|
- -Users-name-projects-my-app-src -> ['my', 'app', 'src']
|
|
365
|
-
- -Users-name-Code-
|
|
365
|
+
- -Users-name-Code-example-project -> ['example', 'project']
|
|
366
366
|
|
|
367
367
|
Args:
|
|
368
368
|
path: Path in any format
|
package/mcp-server/src/server.py
CHANGED
|
@@ -81,15 +81,23 @@ def initialize_embeddings():
|
|
|
81
81
|
print(f"[ERROR] Failed to initialize embeddings: {e}")
|
|
82
82
|
return False
|
|
83
83
|
|
|
84
|
-
# Debug environment loading
|
|
85
|
-
|
|
86
|
-
|
|
87
|
-
|
|
88
|
-
print(f"[
|
|
89
|
-
print(f"[
|
|
90
|
-
print(f"[
|
|
91
|
-
print(f"[
|
|
92
|
-
print(f"[DEBUG]
|
|
84
|
+
# Debug environment loading and startup
|
|
85
|
+
import sys
|
|
86
|
+
import datetime as dt
|
|
87
|
+
startup_time = dt.datetime.now().isoformat()
|
|
88
|
+
print(f"[STARTUP] MCP Server starting at {startup_time}", file=sys.stderr)
|
|
89
|
+
print(f"[STARTUP] Python: {sys.version}", file=sys.stderr)
|
|
90
|
+
print(f"[STARTUP] Working directory: {os.getcwd()}", file=sys.stderr)
|
|
91
|
+
print(f"[STARTUP] Script location: {__file__}", file=sys.stderr)
|
|
92
|
+
print(f"[DEBUG] Environment variables loaded:", file=sys.stderr)
|
|
93
|
+
print(f"[DEBUG] QDRANT_URL: {QDRANT_URL}", file=sys.stderr)
|
|
94
|
+
print(f"[DEBUG] ENABLE_MEMORY_DECAY: {ENABLE_MEMORY_DECAY}", file=sys.stderr)
|
|
95
|
+
print(f"[DEBUG] USE_NATIVE_DECAY: {USE_NATIVE_DECAY}", file=sys.stderr)
|
|
96
|
+
print(f"[DEBUG] DECAY_WEIGHT: {DECAY_WEIGHT}", file=sys.stderr)
|
|
97
|
+
print(f"[DEBUG] DECAY_SCALE_DAYS: {DECAY_SCALE_DAYS}", file=sys.stderr)
|
|
98
|
+
print(f"[DEBUG] PREFER_LOCAL_EMBEDDINGS: {PREFER_LOCAL_EMBEDDINGS}", file=sys.stderr)
|
|
99
|
+
print(f"[DEBUG] EMBEDDING_MODEL: {EMBEDDING_MODEL}", file=sys.stderr)
|
|
100
|
+
print(f"[DEBUG] env_path: {env_path}", file=sys.stderr)
|
|
93
101
|
|
|
94
102
|
|
|
95
103
|
class SearchResult(BaseModel):
|
|
@@ -259,31 +267,31 @@ async def update_indexing_status(cache_ttl: int = 5):
|
|
|
259
267
|
|
|
260
268
|
# Convert set to list for compatibility
|
|
261
269
|
imported_files_list = list(all_imported_files)
|
|
262
|
-
|
|
263
|
-
|
|
264
|
-
|
|
265
|
-
|
|
266
|
-
|
|
267
|
-
|
|
268
|
-
|
|
269
|
-
|
|
270
|
-
|
|
271
|
-
|
|
272
|
-
|
|
273
|
-
|
|
274
|
-
|
|
275
|
-
|
|
276
|
-
|
|
277
|
-
|
|
278
|
-
|
|
279
|
-
|
|
280
|
-
|
|
281
|
-
|
|
282
|
-
|
|
283
|
-
|
|
284
|
-
|
|
285
|
-
|
|
286
|
-
|
|
270
|
+
|
|
271
|
+
# Count files that have been imported
|
|
272
|
+
for file_path in jsonl_files:
|
|
273
|
+
# Normalize the current file path for consistent comparison
|
|
274
|
+
normalized_file = normalize_path(str(file_path))
|
|
275
|
+
|
|
276
|
+
# Try multiple path formats to match Docker's state file
|
|
277
|
+
file_str = str(file_path).replace(str(Path.home()), "/logs").replace("\\", "/")
|
|
278
|
+
# Also try without .claude/projects prefix (Docker mounts directly)
|
|
279
|
+
file_str_alt = file_str.replace("/.claude/projects", "")
|
|
280
|
+
|
|
281
|
+
# Normalize alternative paths as well
|
|
282
|
+
normalized_alt = normalize_path(file_str)
|
|
283
|
+
normalized_alt2 = normalize_path(file_str_alt)
|
|
284
|
+
|
|
285
|
+
# Check if file is in imported_files list (fully imported)
|
|
286
|
+
if normalized_file in imported_files_list or normalized_alt in imported_files_list or normalized_alt2 in imported_files_list:
|
|
287
|
+
indexed_files += 1
|
|
288
|
+
# Or if it has metadata with position > 0 (partially imported)
|
|
289
|
+
elif normalized_file in file_metadata and file_metadata[normalized_file].get("position", 0) > 0:
|
|
290
|
+
indexed_files += 1
|
|
291
|
+
elif normalized_alt in file_metadata and file_metadata[normalized_alt].get("position", 0) > 0:
|
|
292
|
+
indexed_files += 1
|
|
293
|
+
elif normalized_alt2 in file_metadata and file_metadata[normalized_alt2].get("position", 0) > 0:
|
|
294
|
+
indexed_files += 1
|
|
287
295
|
|
|
288
296
|
# Update status
|
|
289
297
|
indexing_status["last_check"] = current_time
|
|
@@ -1519,4 +1527,8 @@ if __name__ == "__main__":
|
|
|
1519
1527
|
sys.exit(0)
|
|
1520
1528
|
|
|
1521
1529
|
# Normal MCP server operation
|
|
1530
|
+
print(f"[STARTUP] Starting FastMCP server in stdio mode...", file=sys.stderr)
|
|
1531
|
+
print(f"[STARTUP] Server name: {mcp.name}", file=sys.stderr)
|
|
1532
|
+
print(f"[STARTUP] Calling mcp.run()...", file=sys.stderr)
|
|
1522
1533
|
mcp.run()
|
|
1534
|
+
print(f"[STARTUP] Server exited normally", file=sys.stderr)
|
package/package.json
CHANGED
|
@@ -0,0 +1,124 @@
|
|
|
1
|
+
#!/usr/bin/env python3
|
|
2
|
+
"""
|
|
3
|
+
Quick import script for current project's latest conversations.
|
|
4
|
+
Designed for PreCompact hook integration - targets <10 second imports.
|
|
5
|
+
"""
|
|
6
|
+
|
|
7
|
+
import os
|
|
8
|
+
import sys
|
|
9
|
+
import json
|
|
10
|
+
import subprocess
|
|
11
|
+
from datetime import datetime, timedelta
|
|
12
|
+
from pathlib import Path
|
|
13
|
+
import logging
|
|
14
|
+
|
|
15
|
+
# Configuration
|
|
16
|
+
LOGS_DIR = os.getenv("LOGS_DIR", os.path.expanduser("~/.claude/projects"))
|
|
17
|
+
STATE_FILE = os.getenv("STATE_FILE", os.path.expanduser("~/.claude-self-reflect-state.json"))
|
|
18
|
+
HOURS_BACK = int(os.getenv("IMPORT_HOURS_BACK", "2")) # Only import last 2 hours by default
|
|
19
|
+
|
|
20
|
+
# Set up logging
|
|
21
|
+
logging.basicConfig(
|
|
22
|
+
level=logging.INFO,
|
|
23
|
+
format='%(asctime)s - %(levelname)s - %(message)s'
|
|
24
|
+
)
|
|
25
|
+
logger = logging.getLogger(__name__)
|
|
26
|
+
|
|
27
|
+
def load_state():
|
|
28
|
+
"""Load import state from file."""
|
|
29
|
+
if os.path.exists(STATE_FILE):
|
|
30
|
+
try:
|
|
31
|
+
with open(STATE_FILE, 'r') as f:
|
|
32
|
+
return json.load(f)
|
|
33
|
+
except:
|
|
34
|
+
return {}
|
|
35
|
+
return {}
|
|
36
|
+
|
|
37
|
+
def save_state(state):
|
|
38
|
+
"""Save import state to file."""
|
|
39
|
+
os.makedirs(os.path.dirname(STATE_FILE), exist_ok=True)
|
|
40
|
+
with open(STATE_FILE, 'w') as f:
|
|
41
|
+
json.dump(state, f, indent=2)
|
|
42
|
+
|
|
43
|
+
def get_project_from_cwd():
|
|
44
|
+
"""Detect project from current working directory."""
|
|
45
|
+
cwd = os.getcwd()
|
|
46
|
+
# Convert path to project name format used in logs
|
|
47
|
+
# Claude logs use format: -Users-username-path-to-project
|
|
48
|
+
project_name = cwd.replace('/', '-')
|
|
49
|
+
# Keep the leading dash as that's how Claude stores it
|
|
50
|
+
if not project_name.startswith('-'):
|
|
51
|
+
project_name = '-' + project_name
|
|
52
|
+
return project_name
|
|
53
|
+
|
|
54
|
+
def get_recent_files(project_path: Path, hours_back: int):
|
|
55
|
+
"""Get JSONL files modified in the last N hours."""
|
|
56
|
+
cutoff_time = datetime.now() - timedelta(hours=hours_back)
|
|
57
|
+
recent_files = []
|
|
58
|
+
|
|
59
|
+
for jsonl_file in project_path.glob("*.jsonl"):
|
|
60
|
+
mtime = datetime.fromtimestamp(jsonl_file.stat().st_mtime)
|
|
61
|
+
if mtime > cutoff_time:
|
|
62
|
+
recent_files.append(jsonl_file)
|
|
63
|
+
|
|
64
|
+
return sorted(recent_files, key=lambda f: f.stat().st_mtime, reverse=True)
|
|
65
|
+
|
|
66
|
+
def main():
|
|
67
|
+
"""Main quick import function."""
|
|
68
|
+
start_time = datetime.now()
|
|
69
|
+
|
|
70
|
+
# Detect current project
|
|
71
|
+
project_name = get_project_from_cwd()
|
|
72
|
+
project_path = Path(LOGS_DIR) / project_name
|
|
73
|
+
|
|
74
|
+
if not project_path.exists():
|
|
75
|
+
logger.warning(f"Project logs not found: {project_path}")
|
|
76
|
+
logger.info("Make sure you're in a project directory with Claude conversations.")
|
|
77
|
+
return
|
|
78
|
+
|
|
79
|
+
logger.info(f"Quick importing latest conversations for: {project_name}")
|
|
80
|
+
|
|
81
|
+
# Get recent files
|
|
82
|
+
recent_files = get_recent_files(project_path, HOURS_BACK)
|
|
83
|
+
logger.info(f"Found {len(recent_files)} files modified in last {HOURS_BACK} hours")
|
|
84
|
+
|
|
85
|
+
if not recent_files:
|
|
86
|
+
logger.info("No recent conversations to import")
|
|
87
|
+
return
|
|
88
|
+
|
|
89
|
+
# For now, just call the unified importer with the specific project
|
|
90
|
+
# This is a temporary solution until we implement incremental imports
|
|
91
|
+
script_dir = os.path.dirname(os.path.abspath(__file__))
|
|
92
|
+
unified_script = os.path.join(script_dir, "import-conversations-unified.py")
|
|
93
|
+
|
|
94
|
+
# Set environment to only process this project
|
|
95
|
+
env = os.environ.copy()
|
|
96
|
+
env['LOGS_DIR'] = str(project_path.parent)
|
|
97
|
+
env['IMPORT_PROJECT'] = project_name
|
|
98
|
+
|
|
99
|
+
try:
|
|
100
|
+
# Run the unified importer for just this project
|
|
101
|
+
result = subprocess.run(
|
|
102
|
+
[sys.executable, unified_script],
|
|
103
|
+
env=env,
|
|
104
|
+
capture_output=True,
|
|
105
|
+
text=True,
|
|
106
|
+
timeout=60 # 60 second timeout
|
|
107
|
+
)
|
|
108
|
+
|
|
109
|
+
if result.returncode == 0:
|
|
110
|
+
logger.info("Quick import completed successfully")
|
|
111
|
+
else:
|
|
112
|
+
logger.error(f"Import failed: {result.stderr}")
|
|
113
|
+
|
|
114
|
+
except subprocess.TimeoutExpired:
|
|
115
|
+
logger.warning("Import timed out after 60 seconds")
|
|
116
|
+
except Exception as e:
|
|
117
|
+
logger.error(f"Error during import: {e}")
|
|
118
|
+
|
|
119
|
+
# Report timing
|
|
120
|
+
elapsed = (datetime.now() - start_time).total_seconds()
|
|
121
|
+
logger.info(f"Quick import completed in {elapsed:.1f} seconds")
|
|
122
|
+
|
|
123
|
+
if __name__ == "__main__":
|
|
124
|
+
main()
|
|
@@ -0,0 +1,171 @@
|
|
|
1
|
+
#!/usr/bin/env python3
|
|
2
|
+
"""
|
|
3
|
+
Import old format JSONL files from Claude conversations.
|
|
4
|
+
These files have a different structure with type/summary fields instead of messages.
|
|
5
|
+
"""
|
|
6
|
+
|
|
7
|
+
import json
|
|
8
|
+
import sys
|
|
9
|
+
from pathlib import Path
|
|
10
|
+
import hashlib
|
|
11
|
+
import uuid
|
|
12
|
+
from datetime import datetime
|
|
13
|
+
from qdrant_client import QdrantClient
|
|
14
|
+
from qdrant_client.models import Distance, VectorParams, PointStruct
|
|
15
|
+
from fastembed import TextEmbedding
|
|
16
|
+
import logging
|
|
17
|
+
|
|
18
|
+
logging.basicConfig(level=logging.INFO, format='%(asctime)s - %(levelname)s - %(message)s')
|
|
19
|
+
logger = logging.getLogger(__name__)
|
|
20
|
+
|
|
21
|
+
def import_old_format_project(project_dir: Path, project_path: str = None):
|
|
22
|
+
"""Import old format JSONL files from a project directory."""
|
|
23
|
+
|
|
24
|
+
# Initialize
|
|
25
|
+
client = QdrantClient(url='http://localhost:6333')
|
|
26
|
+
model = TextEmbedding(model_name='sentence-transformers/all-MiniLM-L6-v2', max_length=512)
|
|
27
|
+
|
|
28
|
+
# Determine project path from directory name if not provided
|
|
29
|
+
if not project_path:
|
|
30
|
+
# Convert -Users-username-projects-projectname back to path
|
|
31
|
+
dir_name = project_dir.name
|
|
32
|
+
project_path = '/' + dir_name.strip('-').replace('-', '/')
|
|
33
|
+
|
|
34
|
+
# Create collection name
|
|
35
|
+
project_hash = hashlib.md5(project_path.encode()).hexdigest()[:8]
|
|
36
|
+
collection_name = f'conv_{project_hash}_local'
|
|
37
|
+
|
|
38
|
+
logger.info(f'Project: {project_path}')
|
|
39
|
+
logger.info(f'Collection: {collection_name}')
|
|
40
|
+
|
|
41
|
+
# Create collection if needed
|
|
42
|
+
try:
|
|
43
|
+
client.get_collection(collection_name)
|
|
44
|
+
logger.info('Collection exists')
|
|
45
|
+
except:
|
|
46
|
+
client.create_collection(
|
|
47
|
+
collection_name=collection_name,
|
|
48
|
+
vectors_config=VectorParams(size=384, distance=Distance.COSINE)
|
|
49
|
+
)
|
|
50
|
+
logger.info('Created collection')
|
|
51
|
+
|
|
52
|
+
# Process all JSONL files
|
|
53
|
+
jsonl_files = list(project_dir.glob('*.jsonl'))
|
|
54
|
+
logger.info(f'Found {len(jsonl_files)} files to import')
|
|
55
|
+
|
|
56
|
+
total_points = 0
|
|
57
|
+
for file_path in jsonl_files:
|
|
58
|
+
logger.info(f'Processing {file_path.name}...')
|
|
59
|
+
points_batch = []
|
|
60
|
+
|
|
61
|
+
with open(file_path, 'r', encoding='utf-8') as f:
|
|
62
|
+
conversation_text = []
|
|
63
|
+
file_timestamp = file_path.stat().st_mtime
|
|
64
|
+
|
|
65
|
+
for line_num, line in enumerate(f, 1):
|
|
66
|
+
try:
|
|
67
|
+
data = json.loads(line)
|
|
68
|
+
msg_type = data.get('type', '')
|
|
69
|
+
|
|
70
|
+
# Extract text content based on type
|
|
71
|
+
content = None
|
|
72
|
+
if msg_type == 'summary' and data.get('summary'):
|
|
73
|
+
content = f"[Conversation Summary] {data['summary']}"
|
|
74
|
+
elif msg_type == 'user' and data.get('summary'):
|
|
75
|
+
content = f"User: {data['summary']}"
|
|
76
|
+
elif msg_type == 'assistant' and data.get('summary'):
|
|
77
|
+
content = f"Assistant: {data['summary']}"
|
|
78
|
+
elif msg_type in ['user', 'assistant']:
|
|
79
|
+
# Try to get content from other fields
|
|
80
|
+
if 'content' in data:
|
|
81
|
+
content = f"{msg_type.title()}: {data['content']}"
|
|
82
|
+
elif 'text' in data:
|
|
83
|
+
content = f"{msg_type.title()}: {data['text']}"
|
|
84
|
+
|
|
85
|
+
if content:
|
|
86
|
+
conversation_text.append(content)
|
|
87
|
+
|
|
88
|
+
# Create chunks every 5 messages or at end
|
|
89
|
+
if len(conversation_text) >= 5:
|
|
90
|
+
chunk_text = '\n\n'.join(conversation_text)
|
|
91
|
+
if chunk_text.strip():
|
|
92
|
+
# Generate embedding
|
|
93
|
+
embedding = list(model.embed([chunk_text[:2000]]))[0] # Limit to 2000 chars
|
|
94
|
+
|
|
95
|
+
point = PointStruct(
|
|
96
|
+
id=str(uuid.uuid4()),
|
|
97
|
+
vector=embedding.tolist(),
|
|
98
|
+
payload={
|
|
99
|
+
'content': chunk_text[:1000], # Store first 1000 chars
|
|
100
|
+
'full_content': chunk_text[:4000], # Store more for context
|
|
101
|
+
'project_path': project_path,
|
|
102
|
+
'file_path': str(file_path),
|
|
103
|
+
'file_name': file_path.name,
|
|
104
|
+
'conversation_id': file_path.stem,
|
|
105
|
+
'chunk_index': len(points_batch),
|
|
106
|
+
'timestamp': file_timestamp,
|
|
107
|
+
'type': 'conversation_chunk'
|
|
108
|
+
}
|
|
109
|
+
)
|
|
110
|
+
points_batch.append(point)
|
|
111
|
+
conversation_text = []
|
|
112
|
+
|
|
113
|
+
except json.JSONDecodeError:
|
|
114
|
+
logger.warning(f'Invalid JSON at line {line_num} in {file_path.name}')
|
|
115
|
+
except Exception as e:
|
|
116
|
+
logger.warning(f'Error processing line {line_num}: {e}')
|
|
117
|
+
|
|
118
|
+
# Handle remaining text
|
|
119
|
+
if conversation_text:
|
|
120
|
+
chunk_text = '\n\n'.join(conversation_text)
|
|
121
|
+
if chunk_text.strip():
|
|
122
|
+
embedding = list(model.embed([chunk_text[:2000]]))[0]
|
|
123
|
+
|
|
124
|
+
point = PointStruct(
|
|
125
|
+
id=str(uuid.uuid4()),
|
|
126
|
+
vector=embedding.tolist(),
|
|
127
|
+
payload={
|
|
128
|
+
'content': chunk_text[:1000],
|
|
129
|
+
'full_content': chunk_text[:4000],
|
|
130
|
+
'project_path': project_path,
|
|
131
|
+
'file_path': str(file_path),
|
|
132
|
+
'file_name': file_path.name,
|
|
133
|
+
'conversation_id': file_path.stem,
|
|
134
|
+
'chunk_index': len(points_batch),
|
|
135
|
+
'timestamp': file_timestamp,
|
|
136
|
+
'type': 'conversation_chunk'
|
|
137
|
+
}
|
|
138
|
+
)
|
|
139
|
+
points_batch.append(point)
|
|
140
|
+
|
|
141
|
+
# Upload batch
|
|
142
|
+
if points_batch:
|
|
143
|
+
client.upsert(collection_name=collection_name, points=points_batch)
|
|
144
|
+
logger.info(f' Uploaded {len(points_batch)} chunks from {file_path.name}')
|
|
145
|
+
total_points += len(points_batch)
|
|
146
|
+
|
|
147
|
+
# Verify
|
|
148
|
+
info = client.get_collection(collection_name)
|
|
149
|
+
logger.info(f'\nImport complete!')
|
|
150
|
+
logger.info(f'Collection {collection_name} now has {info.points_count} points')
|
|
151
|
+
logger.info(f'Added {total_points} new points in this import')
|
|
152
|
+
|
|
153
|
+
return collection_name, total_points
|
|
154
|
+
|
|
155
|
+
def main():
|
|
156
|
+
if len(sys.argv) < 2:
|
|
157
|
+
print("Usage: python import-old-format.py <project-directory> [project-path]")
|
|
158
|
+
print("Example: python import-old-format.py ~/.claude/projects/-Users-me-projects-myapp /Users/me/projects/myapp")
|
|
159
|
+
sys.exit(1)
|
|
160
|
+
|
|
161
|
+
project_dir = Path(sys.argv[1]).expanduser()
|
|
162
|
+
project_path = sys.argv[2] if len(sys.argv) > 2 else None
|
|
163
|
+
|
|
164
|
+
if not project_dir.exists():
|
|
165
|
+
print(f"Error: Directory {project_dir} does not exist")
|
|
166
|
+
sys.exit(1)
|
|
167
|
+
|
|
168
|
+
import_old_format_project(project_dir, project_path)
|
|
169
|
+
|
|
170
|
+
if __name__ == "__main__":
|
|
171
|
+
main()
|