Structured Outputs
GoingMerry supports enforcing structured JSON output formats. By passing a JSON Schema, you can guarantee that the model's completion response conforms to a specific data format.
Tool Calling
Tool calling allows GoingMerry models to invoke external tools or code routines. By providing function definitions in the request, the model decides when to output a structured JSON tool call instead of natural language.
Vision
GoingMerry supports vision capabilities using multimodal models (such as llava, bakllava, or standard vision configurations). You can submit images alongside text prompts to perform OCR, image captioning, and visual QA.
Reasoning & Thinking
For reasoning-capable models (such as deepseek-r1 or thinking configurations), GoingMerry manages the extraction and rendering of internal thinking cycles.
Web Search
Web search allows model completions to query search indexes dynamically to retrieve real-time facts before generating answers.
Embeddings
GoingMerry supports generating high-dimensional vector representations of text. These vector embeddings are critical for RAG (Retrieval-Augmented Generation) systems, PGVector storage, semantic search, and document classification.
Streaming
Streaming allows GoingMerry to output tokens chunk-by-chunk as they are generated by the model. This significantly lowers perceived latency and improves the user experience.