Server Status
Real-time overview of your local ChatSorter instance
—
Facts stored
—
Summaries
—
Chat sessions
—
Queue depth
—
Storage used
—
Uptime
API Key
Loading…
Use this key as
Authorization: Bearer <key> in all API requests to http://localhost:5000Quick Test
Send a test message to verify everything is working end-to-end.
Memory Tuning
Controls how memories are formed, stored and retrieved
Retrieval
Max Memories Returned 10
How many memories come back per search / chat context injection (top_k).
Importance Threshold 0.0
Memories below this score are excluded from results. 0 = return everything.
Recency Decay (hours) 12
Half-life of recency score. Lower = recent memories ranked higher faster.
Similarity Weight 3.0
How much semantic similarity influences ranking vs importance and recency.
Importance Weight 2.0
Recency Weight 1.5
Summarization
Messages per Summary 5
How many messages accumulate before a summary is generated. Lower = more frequent summaries, more model calls.
Min sentences 2
Max sentences 4
Retention
Shred after days 365
Low-importance memories older than this are removed by /admin/shred-history.
Keep summaries above 8.0
Keep facts above 0.8
Memory Scope
Cross-Chat Memory
When ON, facts from any chat are visible across all conversations. When OFF (default), each chat has isolated memory.
Model Settings
Configure which Ollama models to use. Changes require a server restart.
Ollama Connection
Ollama URL
Where your Ollama instance is running.
Model Selection
Summarizer model
Used to generate memory summaries. Recommended: qwen2.5:3b (fast) or qwen2.5:7b (better quality).
Extractor model
Used to extract structured facts. Leave blank to use the same model as the summarizer.
Embedder model
Used for semantic search. Must be an embedding model.
Generation Parameters
Summary temperature 0.2
Lower = more consistent. Higher = more creative.
Extraction temperature 0.1
Keep low for consistent fact extraction.
Max output tokens 360
Context window 2048
Keep-alive seconds 300
How long Ollama keeps the model loaded in VRAM after the last request. -1 = forever.
Server
Port
Data directory
⚠ These changes require a server restart to take effect.
Custom Prompts
Override the default summarization and extraction prompts. Leave blank to use built-in defaults.
Summarization Prompt
This prompt is sent to the model when generating a memory summary from a conversation batch. Use
{messages} as a placeholder for the conversation text.Extraction Prompt
This prompt is sent to the model when extracting structured facts from a message. Use
{message} as a placeholder.Memory Viewer
Browse and manage stored facts and summaries
Filter
User ID
Chat ID
Facts 0
| Chat | Type / Value | Score | Age |
|---|---|---|---|
| Load memories above | |||
Summaries 0
Load memories above
Backup & Restore
Export all memory data to a JSON file, or restore from a previous backup
Export
Downloads all stored facts and summaries as a single JSON file. Safe to do while the server is running.
User (optional)
Import / Restore
Upload a backup file exported from this instance or another ChatSorter Local installation.
Drop backup JSON here or click to browse
No file selectedMerge with existing data
ON = append imported memories to existing ones. OFF = overwrite.
Server Logs
Last 200 lines from the server log files
Log output
Click Refresh to load logs…
License
Your ChatSorter Local license status and verification
License Status
| Status | — |
| Plan | — |
| Expires | — |
| License key | — |
| Live verification | — |
| Server version | — |
Restart Server
After changing model, port, or data directory settings, restart the server to apply them. The API will be unavailable for ~2 seconds.