FonaTech/Clouds-Coder

Clouds Coder – What it is

Clouds Coder is a Python‑based runtime that lets you run command‑line tools (the CLI execution plane) on a local or cloud host while interacting with the process through a rich browser‑based IDE (the Web user plane). It is built around general‑purpose AI agents that can drive the CLI, and it adds a LAN‑wide collaboration mode so multiple people and their private agents can work on the same project without stepping on each other’s toes.

The core idea is to separate the heavy‑lifting execution (e.g., git, make, docker, custom scripts) from the user interface, giving you:

• observable, replayable runs (full logs, debug console, terminal output);
• a versioned workspace with file‑level history;
• a “prompt enhancer” that expands a user’s request into a richer, depth‑budgeted plan for the agent;
• fine‑grained security – every tool invocation must be approved by an admin before it can run in a given workspace.

---

Key components (as described in the README)

Component	What it does
Runtime (CLI execution plane)	Executes arbitrary commands, runs Python debugpy or pdb for debugging, and feeds results back to the UI.
Browser IDE	Monaco‑based editor, file explorer, terminal, debug console, and panels for Agent chat, task blackboard, and diagnostics.
Liquid Kernel	Immutable control plane that stores signed tool‑policy, prompt‑policy, and lifecycle hooks (before_run, before_round, after_tool_results). Handles canary roll‑outs and roll‑backs of kernel versions.
Collaboration Mode (LAN)	A separate listener (P+7 by default) that hosts a governed project workspace. Approved devices join via a password + device‑key flow, get 24 h HttpOnly sessions, and can see revisioned files, presence, and a shared task blackboard.
Agent orchestration	Within a session agents can take roles (manager, explorer, developer, reviewer). Across a project a coordinator agent slices work and contributor agents execute non‑overlapping slices.
Prompt Enhancer	One‑shot or persistent prompting with configurable depth budgets (Low → XHigh). It adds workspace awareness and optional skill orchestration before sending the request to the LLM.
Security & Trust Boundary	Admins approve devices, define which MCP (model‑controlled‑process) commands are allowed, and sign a registry stored in SQLite. Any change to the approved configuration invalidates existing approvals.
Revision‑aware write protocol	Text edits use operational transforms; binary/whole‑file writes use optimistic concurrency; shell/agent processes acquire a mutation lease to guarantee atomic multi‑file updates.

---

Typical workflow

1. Start the runtime – python Clouds_Coder.py --host 0.0.0.0 --port 8080 (runtime on 8080, IDE on 8085, collaboration on 8087).
2. Open the IDE – navigate to http://127.0.0.1:8085. The IDE shows the session workspace, editor, terminal, and an Agent chat panel.
3. Ask the agent – type a high‑level goal (e.g., “set up a Flask project with Docker”). The Prompt Enhancer expands the request, the agent plans steps, and the runtime executes the CLI commands while streaming output to the IDE.
4. Collaborate – enable --enable_collaboration. Invite teammates via the admin UI (/admin → Collaboration). After device approval, teammates open the project URL (P+7) and see the same revisioned files, a shared blackboard, and each other’s presence.
5. Resolve conflicts – if two agents try to modify the same file, the system creates candidate branches, freezes the baseline, and forces a review by two distinct members before merging.
6. Audit & rollback – all actions are signed and stored in an immutable SQLite audit chain. Admins can roll back a kernel version or restore a previous project snapshot.

---

Who might use it?

• AI‑agent developers who need a reproducible environment to test agents that drive CLIs.
• Teams of developers who want a low‑friction, LAN‑only collaborative IDE that keeps every command observable and auditable.
• Educators / labs teaching “Vibe Coding” – the practice of guiding an AI through a coding task while watching the terminal and file changes live.
• Security‑conscious organizations that require explicit approval for any tool invocation and want a full audit trail.

---

Installation & quick start (from the README)

pip install clouds-coder   # pulls the latest PyPI release
# start a runtime on port 8080 (IDE will be on 8085, collaboration on 8087)
python Clouds_Coder.py \
  --host 0.0.0.0 \
  --port 8080 \
  --enable_collaboration \
  --collab_host 0.0.0.0 \
  --collab_port 8087

• Open http://127.0.0.1:8085 for the IDE.
• Open http://127.0.0.1:8080/admin to create a project, set a password, and approve devices.
• For production LAN use, supply --collab_tls_cert/--collab_tls_key or place an HTTPS reverse proxy in front.

---

Where to learn more

• Changelogs – detailed feature histories are kept in log/CHANGELOG‑*.md (e.g., the 2026‑09‑12 “Liquid Kernel & Scalable Session Runtime” changelog).
• LLM Config Template – LLM.config.json shows how to plug in server‑side LLM profiles, skills, and RAG resources.
• Architecture diagrams – the README includes Mermaid flowcharts that illustrate the runtime, control plane, and collaboration boundaries.

---

Bottom line

Clouds Coder is a real, production‑grade platform that couples an AI‑driven CLI executor with a web‑based IDE and a governed LAN collaboration layer. Its focus is on making complex command‑line tasks observable, reproducible, and safe for multiple human and agent participants.