🦙 Bulk Prompt Ollama Improver

Improve each prompt with a local Ollama model (official ollama library), optionally trimming any chat wrapper the model adds. Use the 🔄 Reconnect button to load the model list.

By sifatrads·Created 3 months ago·Updated 3 months ago· 0

result
context
meta

◄instructionYou are an expert prompt engineer for image generation. Rewrite the following prompt to be more vivid and detailed. Output ONLY the improved prompt, with no preamble or explanation.►

◄urlhttp://127.0.0.1:11434►

◄model▾►

◄keep_alive_minutes5►

◄timeout120►

◄enabledtrue►

◄trim_outputtrue►

◄positive—►

◄context—►

CategoryBulkPrompt

Inputs (9)

Name	Type	Default	Description
instruction	STRING	You are an expert prompt engineer for image generation. Rewrite the following prompt to be more vivid and detailed. Output ONLY the improved prompt, with no preamble or explanation.	System prompt — tell the model what to do with the prompt.
url	STRING	http://127.0.0.1:11434	Ollama server URL. Change this for a custom host/port (e.g. http://192.168.1.50:11434), then click 🔄 Reconnect.
model	COMBO		Pick a model (auto-loaded from the Ollama server at the default URL). After changing url or pulling new models, click 🔄 Reconnect to refresh the list.
keep_alive_minutes	INT	5-1–1440	Minutes Ollama keeps the model loaded after inference (-1 = keep forever, 0 = unload immediately).
timeout	INT	1205–3600	Max seconds to wait for the response. Large models (e.g. 14B) need more — the first run also loads the model into VRAM. Use 300+ if you hit timeouts.
enabled	BOOLEAN	true	If off, the prompt passes through unchanged (no Ollama call).
trim_output	BOOLEAN	true	Keep only the prompt: strip any chat wrapper the model adds (preambles, '---'/code-fence/'Prompt:' headers, trailing offers). Turn off to use the raw reply.
positive	STRING		—
contextopt	STRING		Optional Ollama context (comma-separated ints) for multi-turn continuity.

Outputs (3)

Name	Type	Description
result	STRING	—
context	STRING	—
meta	STRING	—