# yannicksiewe/skyops-coder architecture

> Architecture diagram of the GitHub repository yannicksiewe/skyops-coder, generated by GitDiagram from its file tree, README and a sample of source files (last updated 2026-10-02). It is an AI-made overview: check details against the source.

- Interactive diagram: https://gitdiagram.com/yannicksiewe/skyops-coder
- Repository: https://github.com/yannicksiewe/skyops-coder

## Overview

skyops-coder is a self-hosted engineering AI platform: team tools send chat, code, image, or agent requests through compatible model endpoints, while separate workflows provide fine-tuning and repository automation. Its documented runtime also includes gateway accounting and tracing, monitoring, and encrypted backups. The README supports the major deployment relationships; sampled source confirms agent orchestration, shared LLM/GitHub helpers, diff parsing, MCP requests, and tool-call parsing. Other operational links are shown only where documented, not inferred from configuration names.

Main entry points: [`serving/vllm-chat.service`](https://github.com/yannicksiewe/skyops-coder/blob/main/serving/vllm-chat.service) (Chat model), [`ops/gateway/compose.yml`](https://github.com/yannicksiewe/skyops-coder/blob/main/ops/gateway/compose.yml) (LLM gateway), [`.github/workflows/ai-task.yml`](https://github.com/yannicksiewe/skyops-coder/blob/main/.github/workflows/ai-task.yml) (PR and deploy workflows), [`ops/monitoring/compose.yml`](https://github.com/yannicksiewe/skyops-coder/blob/main/ops/monitoring/compose.yml) (Prometheus and Grafana)

## Diagram (Mermaid)

```mermaid
flowchart TD

subgraph group_serving["Model serving"]
  node_chat["Chat model<br/>[vllm-chat.service]"]
  node_autocomplete["Autocomplete model"]
  node_vision["Vision model"]
  node_parser["Tool-call parser"]
end

subgraph group_clients["Team interfaces"]
  node_webui["Open WebUI<br/>[webui_config.py]"]
  node_ide["IDE and SDK clients"]
  node_aider["Aider client<br/>[aider.env]"]
  node_mcp["MCP tools<br/>[skyops_mcp.py]"]
end

subgraph group_agents["Repository agents"]
  node_review["Code review agent<br/>[review.py]"]
  node_architecture["Architecture review agent<br/>[arch_review.py]"]
  node_issueagent["Issue-to-PR agent<br/>[issue_to_pr.sh]"]
  node_agenthelpers["Agent LLM and GitHub<br/>[common.py]"]
  node_diffparse["Diff parser<br/>[diffparse.py]"]
end

subgraph group_training["Model training"]
  node_train["LoRA fine-tuning<br/>[train.py]"]
  node_infer["Adapter inference<br/>[infer.py]"]
end

subgraph group_operations["Platform operations"]
  node_gateway["LLM gateway<br/>[compose.yml]"]
  node_guardrail["Secrets guardrail"]
  node_runner["Self-hosted runner<br/>[install_runner.sh]"]
  node_workflows["PR and deploy workflows<br/>[ai-task.yml]"]
  node_tracing["Request tracing<br/>[compose.yml]"]
  node_monitoring["Prometheus and Grafana<br/>[compose.yml]"]
  node_backups["Encrypted backups<br/>[backup.sh]"]
end

node_engineer(("Engineering team"))
node_browser(("Browser user"))
node_github["GitHub"]

node_engineer -->|"uses"| node_ide
node_engineer -->|"uses"| node_aider
node_engineer -->|"invokes tools"| node_mcp
node_browser -->|"uses"| node_webui
node_ide -->|"requests chat"| node_chat
node_ide -->|"requests completion"| node_autocomplete
node_aider -->|"requests chat"| node_chat
node_webui -->|"connects to"| node_chat
node_webui -->|"connects to"| node_autocomplete
node_mcp -->|"sends API requests"| node_gateway
node_gateway -->|"routes requests"| node_chat
node_gateway -->|"routes requests"| node_vision
node_chat -->|"extracts tool calls"| node_parser
node_workflows -->|"runs on"| node_runner
node_workflows -->|"runs review"| node_review
node_workflows -->|"runs review"| node_architecture
node_review -->|"parses diffs"| node_diffparse
node_review -->|"uses LLM and GitHub"| node_agenthelpers
node_architecture -->|"parses diffs"| node_diffparse
node_architecture -->|"uses LLM and GitHub"| node_agenthelpers
node_issueagent -.->|"creates PR"| node_github
node_train -.->|"produces adapter"| node_infer
node_gateway -->|"applies redaction"| node_guardrail
node_gateway -->|"records traces"| node_tracing
node_monitoring -->|"monitors services"| node_chat
node_monitoring -->|"monitors services"| node_autocomplete
node_monitoring -->|"monitors services"| node_vision

click node_chat "https://github.com/yannicksiewe/skyops-coder/blob/main/serving/vllm-chat.service"
click node_autocomplete "https://github.com/yannicksiewe/skyops-coder/blob/main/serving/vllm-autocomplete.service"
click node_vision "https://github.com/yannicksiewe/skyops-coder/blob/main/serving/vllm-vision.service"
click node_parser "https://github.com/yannicksiewe/skyops-coder/blob/main/serving/coder_tool_parser.py"
click node_webui "https://github.com/yannicksiewe/skyops-coder/blob/main/serving/webui_config.py"
click node_ide "https://github.com/yannicksiewe/skyops-coder/blob/main/clients/continue-config.yaml"
click node_aider "https://github.com/yannicksiewe/skyops-coder/blob/main/clients/aider.env"
click node_mcp "https://github.com/yannicksiewe/skyops-coder/blob/main/clients/mcp/skyops_mcp.py"
click node_gateway "https://github.com/yannicksiewe/skyops-coder/blob/main/ops/gateway/compose.yml"
click node_guardrail "https://github.com/yannicksiewe/skyops-coder/blob/main/ops/gateway/guardrails/secrets_guardrail.py"
click node_review "https://github.com/yannicksiewe/skyops-coder/blob/main/agents/review.py"
click node_architecture "https://github.com/yannicksiewe/skyops-coder/blob/main/agents/arch_review.py"
click node_issueagent "https://github.com/yannicksiewe/skyops-coder/blob/main/agents/issue_to_pr.sh"
click node_agenthelpers "https://github.com/yannicksiewe/skyops-coder/blob/main/agents/common.py"
click node_diffparse "https://github.com/yannicksiewe/skyops-coder/blob/main/agents/diffparse.py"
click node_train "https://github.com/yannicksiewe/skyops-coder/blob/main/training/train.py"
click node_infer "https://github.com/yannicksiewe/skyops-coder/blob/main/training/infer.py"
click node_runner "https://github.com/yannicksiewe/skyops-coder/blob/main/ops/runner/install_runner.sh"
click node_workflows "https://github.com/yannicksiewe/skyops-coder/blob/main/.github/workflows/ai-task.yml"
click node_tracing "https://github.com/yannicksiewe/skyops-coder/blob/main/ops/langfuse/compose.yml"
click node_monitoring "https://github.com/yannicksiewe/skyops-coder/blob/main/ops/monitoring/compose.yml"
click node_backups "https://github.com/yannicksiewe/skyops-coder/blob/main/ops/backup/backup.sh"

classDef toneNeutral fill:#f8fafc,stroke:#334155,stroke-width:1.5px,color:#0f172a
classDef toneBlue fill:#dbeafe,stroke:#2563eb,stroke-width:1.5px,color:#172554
classDef toneAmber fill:#fef3c7,stroke:#d97706,stroke-width:1.5px,color:#78350f
classDef toneMint fill:#dcfce7,stroke:#16a34a,stroke-width:1.5px,color:#14532d
classDef toneRose fill:#ffe4e6,stroke:#e11d48,stroke-width:1.5px,color:#881337
classDef toneIndigo fill:#e0e7ff,stroke:#4f46e5,stroke-width:1.5px,color:#312e81
classDef toneTeal fill:#ccfbf1,stroke:#0f766e,stroke-width:1.5px,color:#134e4a
class node_chat,node_autocomplete,node_vision,node_parser,node_browser toneBlue
class node_webui,node_ide,node_aider,node_mcp toneAmber
class node_review,node_architecture,node_issueagent,node_agenthelpers,node_diffparse toneMint
class node_train,node_infer toneRose
class node_gateway,node_guardrail,node_runner,node_workflows,node_tracing,node_monitoring,node_backups,node_engineer,node_github toneIndigo
```

## Components

### Model serving

- [Chat model](https://github.com/yannicksiewe/skyops-coder/blob/main/serving/vllm-chat.service): `serving/vllm-chat.service`
- [Autocomplete model](https://github.com/yannicksiewe/skyops-coder/blob/main/serving/vllm-autocomplete.service): `serving/vllm-autocomplete.service`
- [Vision model](https://github.com/yannicksiewe/skyops-coder/blob/main/serving/vllm-vision.service): `serving/vllm-vision.service`
- [Tool-call parser](https://github.com/yannicksiewe/skyops-coder/blob/main/serving/coder_tool_parser.py): `serving/coder_tool_parser.py`

### Team interfaces

- [Open WebUI](https://github.com/yannicksiewe/skyops-coder/blob/main/serving/webui_config.py): `serving/webui_config.py`
- [IDE and SDK clients](https://github.com/yannicksiewe/skyops-coder/blob/main/clients/continue-config.yaml): `clients/continue-config.yaml`
- [Aider client](https://github.com/yannicksiewe/skyops-coder/blob/main/clients/aider.env): `clients/aider.env`
- [MCP tools](https://github.com/yannicksiewe/skyops-coder/blob/main/clients/mcp/skyops_mcp.py): `clients/mcp/skyops_mcp.py`

### Repository agents

- [Code review agent](https://github.com/yannicksiewe/skyops-coder/blob/main/agents/review.py): `agents/review.py`
- [Architecture review agent](https://github.com/yannicksiewe/skyops-coder/blob/main/agents/arch_review.py): `agents/arch_review.py`
- [Issue-to-PR agent](https://github.com/yannicksiewe/skyops-coder/blob/main/agents/issue_to_pr.sh): `agents/issue_to_pr.sh`
- [Agent LLM and GitHub](https://github.com/yannicksiewe/skyops-coder/blob/main/agents/common.py): `agents/common.py`
- [Diff parser](https://github.com/yannicksiewe/skyops-coder/blob/main/agents/diffparse.py): `agents/diffparse.py`

### Model training

- [LoRA fine-tuning](https://github.com/yannicksiewe/skyops-coder/blob/main/training/train.py): `training/train.py`
- [Adapter inference](https://github.com/yannicksiewe/skyops-coder/blob/main/training/infer.py): `training/infer.py`

### Platform operations

- [LLM gateway](https://github.com/yannicksiewe/skyops-coder/blob/main/ops/gateway/compose.yml): `ops/gateway/compose.yml`
- [Secrets guardrail](https://github.com/yannicksiewe/skyops-coder/blob/main/ops/gateway/guardrails/secrets_guardrail.py): `ops/gateway/guardrails/secrets_guardrail.py`
- [Self-hosted runner](https://github.com/yannicksiewe/skyops-coder/blob/main/ops/runner/install_runner.sh): `ops/runner/install_runner.sh`
- [PR and deploy workflows](https://github.com/yannicksiewe/skyops-coder/blob/main/.github/workflows/ai-task.yml): `.github/workflows/ai-task.yml`
- [Request tracing](https://github.com/yannicksiewe/skyops-coder/blob/main/ops/langfuse/compose.yml): `ops/langfuse/compose.yml`
- [Prometheus and Grafana](https://github.com/yannicksiewe/skyops-coder/blob/main/ops/monitoring/compose.yml): `ops/monitoring/compose.yml`
- [Encrypted backups](https://github.com/yannicksiewe/skyops-coder/blob/main/ops/backup/backup.sh): `ops/backup/backup.sh`

### Other components

- **Engineering team**
- **Browser user**
- **GitHub**

## Connections

- Engineering team → IDE and SDK clients: uses (evidence: `README.md`)
- Engineering team → Aider client: uses (evidence: `README.md`)
- Engineering team → MCP tools: invokes tools (evidence: `README.md`)
- Browser user → Open WebUI: uses (evidence: `README.md`)
- IDE and SDK clients → Chat model: requests chat (evidence: `README.md`)
- IDE and SDK clients → Autocomplete model: requests completion (evidence: `README.md`)
- Aider client → Chat model: requests chat (evidence: `README.md`)
- Open WebUI → Chat model: connects to (evidence: `README.md`)
- Open WebUI → Autocomplete model: connects to (evidence: `README.md`)
- MCP tools → LLM gateway: sends API requests (evidence: `clients/mcp/skyops_mcp.py`)
- LLM gateway → Chat model: routes requests (evidence: `README.md`)
- LLM gateway → Vision model: routes requests (evidence: `README.md`)
- Chat model → Tool-call parser: extracts tool calls (evidence: `serving/coder_tool_parser.py`)
- PR and deploy workflows → Self-hosted runner: runs on (evidence: `README.md`)
- PR and deploy workflows → Code review agent: runs review (evidence: `README.md`)
- PR and deploy workflows → Architecture review agent: runs review (evidence: `README.md`)
- Code review agent → Diff parser: parses diffs (evidence: `agents/review.py`)
- Code review agent → Agent LLM and GitHub: uses LLM and GitHub (evidence: `agents/review.py`)
- Architecture review agent → Diff parser: parses diffs (evidence: `agents/arch_review.py`)
- Architecture review agent → Agent LLM and GitHub: uses LLM and GitHub (evidence: `agents/arch_review.py`)
- Issue-to-PR agent → GitHub: creates PR
- LoRA fine-tuning → Adapter inference: produces adapter
- LLM gateway → Secrets guardrail: applies redaction (evidence: `README.md`)
- LLM gateway → Request tracing: records traces (evidence: `README.md`)
- Prometheus and Grafana → Chat model: monitors services (evidence: `README.md`)
- Prometheus and Grafana → Autocomplete model: monitors services (evidence: `README.md`)
- Prometheus and Grafana → Vision model: monitors services (evidence: `README.md`)
