Supported Use Cases and Workflows

DevSpeak is designed to support a variety of workflows across the software development lifecycle.

Common Use Cases

1. Rapid Specification Generation

Convert informal requirements to formal technical specs in seconds.

Workflow:

1. Navigate to "Translation Editor."

2. Paste informal requirement into Non-Technical Input.

3. Select Target Audience (e.g., Senior Dev).

4. Select Tech Context (e.g., Backend).

5. Set Verbosity to BALANCED (50) for standard output.

6. Click "TRANSLATE TEXT."

7. Wait for the output to finish generating.

8. Copy/export translated specification.

Expected Outcome: Professional technical specification ready for architectural review.

2. Format-Specific Generation (e.g., Jira Tickets)

Generate structured, actionable tickets for your engineering team.

Workflow:

1. Enter informal requirement.

2. Change OUTPUT FORMAT to "Jira Tickets."

3. Adjust Verbosity based on detail needed.

4. Generate translation.

5. Output structured as Jira-compatible markdown with fields: Summary, Description, Acceptance Criteria, Tasks.

3. Document Translation

Translate entire documents (PDFs, Markdown, TXT) into technical specifications.

Workflow:

1. Navigate to "Translate Files."

2. Drag-and-drop PDF/Markdown/TXT file.

3. Configure same parameters as text translation.

4. System parses document, preserving formatting.

5. Chunking strategy processes large docs in segments.

6. Download translated document in original format.

7. Preview in side-by-side comparison view.

4. Iterative Refinement

Refine and iterate on translations to achieve the perfect output.

Workflow:

1. Initial translation (CONCISE mode).

2. Review in History.

3. Re-run with DETAILED mode for more context.

4. Compare outputs side-by-side.

5. Finalize preferred version for technical documentation.

5. AI Agent Workflows (AgentID + API/CLI/MCP)

Agents sign in with Continue with AgentID (their own verified identity, section: Authentication and Session Management) and then work DevSpeak like any authenticated user — dashboard, API keys, CLI, or MCP. Ten workflows that fit agent loops:

1. Turn a chat request into a build-ready spec (audience + format, e.g. Senior Dev + Technical Spec).

2. Pre-optimize messy stakeholder text (Stage 1) before translating, so typos never reach the spec.

3. Generate Jira-ready tickets with acceptance criteria from one paragraph of intent.

4. Refine iteratively via POST /api/v1/refine — previous output plus feedback, no prompt rewriting.

5. Ground output in the real repo via a Context Project (instructions, assets, connectors) instead of inventing stack details.

6. Emit machine-readable specs (responseFormat: xml) for scaffolders and downstream pipeline steps.

7. Draft API designs (endpoints, schemas, contracts) to implement against directly.

8. Translate inside scripts and CI with the CLI (devspeak translate … --format …, stdin pipes; docs to stdout, diagnostics to stderr).

9. Call /api/v1/translate programmatically with a dsk_live_ key (Settings → API Keys).

10. Keep specs evergreen with Refine — every pass returns a complete standalone replacement, so the agent always holds the current version.