DeepSeek

Ingest codebase via DeepSeek-Coder-V2 API.

Apply custom prompt for security vulnerability scanning.

DeepSeek identifies specific line numbers and provides patches.

Integrate output into CI/CD pipeline for automated blocking.

Switch to DeepSeek-R1 (Reasoner) model.

Input the theorem and known constraints.

Model generates a long Chain-of-Thought (CoT) trace.

Final proof is verified against Lean or formal verification tools.

Load product documentation into a Vector DB.

Use DeepSeek-V3 for RAG retrieval.

Set system instruction to detect input language and respond natively.

Output structured JSON for ticket logging.

Batch upload documents using DeepSeek's context caching.

Execute summarization tasks at $0.14 per million tokens.

Aggregate results into a central BI dashboard.

Utilize Python scripts for concurrent API calls.

Initialize agent with DeepSeek-R1.

Describe the feature request (e.g., 'Add OAuth2 to this FastAPI app').

Agent scans file tree and proposes structural changes.

User reviews and applies the diff provided by the model.

Pass raw text or OCR output to DeepSeek-V3.

Define a Pydantic model for the desired JSON structure.

Instruct the model to extract and validate fields.

Integrate JSON directly into SQL databases.

Leverage DeepSeek-R1's visible CoT traces.

Display the 'thinking' process to the student.

Identify precisely where the student's logic diverged from the correct path.

Provide iterative hints rather than a full solution.