dct-agent 3.0.0

Doraemon Cyber Team — Multi-server Ollama agent

DCT Agent — Doraemon Cyber Team

Multi-server Ollama agent · v3.1 made by baba01hacker

Install

pip install -e .
# module entrypoint (no console script required):
python -m dct

Quick Start

# Register servers
python -m dct add localhost 11434 local
python -m dct add 192.168.1.10 11434 home "home lab box"
python -m dct add 10.0.0.5 11434 vps1 "DigitalOcean"

# Probe all
python -m dct probe

# List all models across servers
python -m dct models

# Launch interactive shell
python -m dct

Shell Commands

Server Management

Command	Description
/servers	List all servers with status, latency, model count
/add <host> <port> [alias] [note]	Register + probe a new server
/remove <alias|#>	Unregister a server
/probe	Probe all servers in parallel
/probe <alias|#>	Detailed probe of one server (all endpoints)
/use <alias|#>	Switch active server

Model Management

Command	Description
/models	Models on active server
/models <alias|#>	Models on specific server
/allmodels	All models across all servers
/model <n>	Switch model
/pull <model>	Pull on active server
/pull <alias|#> <model>	Pull on specific server
/delete <model>	Delete from active server
/show <model>	Model info, system prompt, modelfile

Chat & Session

Command	Description
/clear	Clear history
/history	Turn count + token estimate
/system <prompt>	Set system prompt
/prompts	List built-in system prompt presets
/prompt <name>	Apply a built-in system prompt preset
/copy	Copy transcript to clipboard (fallback prints text)
/save <file>	Save conversation JSON
/load <file>	Resume saved conversation
/status	Full server table

Advanced Agent Control

Command	Description
/rewind or /back	Rewind conversation by 1 turn (drop last user prompt and AI response)
/retry	Rewind conversation and immediately resend the last user prompt
/editai	Open the last AI response in an interactive editor to manually steer the agent
/commit	Generate a conventional Git commit message for staged files using the AI model

Agent Mode

Command	Description
/agent	Toggle agent mode ON/OFF (ON by default on startup)
/agent status	Show current state

Agent mode is active by default. You can disable it on startup using the --no-agent CLI flag. When agent mode is ON, the model autonomously calls tools:

• run_python — execute Python
• run_bash — execute bash
• run_shell — shell command
• read_file — read file (supports <start_line>, <end_line>, and <tail> slicing to save context)
• write_file — write/create file (includes Python syntax validation via ast)
• patch_file — find+replace in file (includes Python syntax validation via ast)
• list_dir — list directory
• tree — directory tree
• fetch_url — fetch a URL
• web_search — DuckDuckGo search
• run_subagent — spawn a sub-agent to delegate a sub-task (accepts <instruction>, optional <model>, optional <system_prompt>)

Direct Tools (no model)

Command	Description
/run python <code>	Execute Python directly
/run bash <code>	Execute bash directly
/read <path>	Read and display file
/write <path>	Write file (interactive)
/fetch <url>	Fetch URL
/search <query>	Web search

Broadcast

Command	Description
/broadcast <message>	Send to ALL online servers simultaneously
/bc <message>	Same

Side Queries (BTW)

Command	Description
/btw <question>	Ask the active model a side question using current context without modifying the session history

Model Router

The router picks the best server automatically:

1. Preferred alias + model (if specified)
2. Fastest online server that has the requested model
3. Any online server (uses its first model)

Failover is automatic — if the active server dies mid-chat, the agent reprobes and jumps to the next available server.

Reliability & Resilience

Built for long-running autonomous tasks, DCT Agent implements professional-grade resilience:

• Context Pruning & Summarization: A sliding window monitors the context token size. When it gets too large (~30k tokens), the agent drops older tool executions and autonomously makes a secondary API call to summarize the dropped interactions. This summary is injected as a persistent memory, preventing token exhaustion while maintaining high-level task awareness.
• Network Retry Backoff: All requests use a robust HTTP session that automatically intercepts transient API errors (429 Too Many Requests, 502 Bad Gateway, 503 Service Unavailable) and retries with an exponential backoff.

Server Registry

Servers persist in ~/.config/dct/servers.json. Add once, use everywhere.

Requirements

rich>=13.0
requests>=2.31
Python>=3.11