Search documentation
Find guides and API references.
Workflow guides
Task-first guides for weekly reviews, bring-your-own-key setup, self-hosted Whisper, and voice notes in SystemSculpt for Obsidian.
Verified workflow scope
These guides are written against the current website docs and plugin surface: chat context chips, Agent Mode approvals, provider settings, recording controls, and the OpenAI-compatible transcription endpoint shape. Re-run the release checklist before changing plugin UI labels.
Outcome: A weekly note drafted from daily notes, unresolved checkboxes, decisions, and blockers.
- Open your daily notes for the review window and attach them to a new SystemSculpt chat.
- Ask Agent Mode to extract incomplete checkboxes, decisions, blockers, and recurring themes.
- Review the proposed weekly note before any file write, then approve only the final note creation.
- Save the chat next to the weekly note so the reasoning trail stays in your vault.
Prompt pattern: ask SystemSculpt to group items by project, preserve source-note links, and produce a final markdown note with sections for wins, carry-forwards, decisions, and next-week priorities. Keep Agent Mode approvals on so file creation is reviewable before it touches the vault.
Outcome: A free plugin setup that uses your own OpenRouter key, Ollama server, or other custom provider.
- Install the plugin for free and open Settings -> SystemSculpt -> Overview & Setup.
- Choose a custom cloud provider such as OpenRouter, or a local HTTP provider such as Ollama.
- Paste the provider key or base URL, add the exact model IDs you want, and run Test connection.
- Use paid SystemSculpt plans only when you want managed models, hosted credits, or Pro convenience features.
License boundary: BYOK chat keeps provider billing with the provider you choose. Paid SystemSculpt plans are for managed-model convenience, hosted credit-metered features, and Pro account surfaces; they do not remove your ability to use custom providers.
Outcome: A local or LAN transcription service that SystemSculpt can call from Obsidian.
- POST multipart/form-data to an OpenAI-compatible /v1/audio/transcriptions endpoint.
- Send the audio file as file, the model name as model, and optional language, prompt, or response_format fields.
- Return JSON with at least a text string. Segment arrays are optional but useful for timestamps.
- Keep the endpoint reachable from the machine running Obsidian, for example http://localhost:9000/v1/audio/transcriptions.
Reference server shape
POST /v1/audio/transcriptions
Content-Type: multipart/form-data
file=<audio bytes>
model=whisper-large-v3
200 OK
{ "text": "transcribed text" }Outcome: A recorded voice note saved as audio, transcribed, and attached to chat as vault context.
- Select your microphone under Settings -> SystemSculpt -> Recording.
- Use the recorder control or Toggle Audio Recorder command to start and stop capture.
- Confirm whether originals should be kept in SystemSculpt/Recordings after transcription.
- Attach the transcript to chat or save it as a note before using it in an agent workflow.
Current recorder behavior: recording starts from the recorder control or command palette, transcription uses the selected SystemSculpt/custom/local provider, and saved transcripts can be reused in chat or weekly-review workflows like any other note.
Before support escalation
Capture the plugin version, provider name, model ID, endpoint URL without secrets, and a screenshot of the failing notice. Then run the troubleshooting guide so support can separate setup mistakes from product bugs.
Open troubleshooting