llm-chat 0.2.1 adds the `keep_alive` input, so a flow can keep an Ollama model resident instead of paying a 30 to 40 second reload after five minutes of idling. llm-embed 0.1.0 is new: text to vectors, with batching as the primary case. A list of texts becomes one request and the result keeps input order. Both signed against the Flemming.AI key and verified before upload; a one-byte change to either bundle is rejected by its signature. SHA256.txt was incomplete: the three 0.2.0 bundles published last time never made it in. Regenerated from the bundles themselves, with every pre-existing line confirmed unchanged first, so the file describes what is actually here rather than part of it. Signed-off-by: Stefan Flemming <stefan.a.flemming@googlemail.com>
1 line
97 B
Standard ML
1 line
97 B
Standard ML
MEUCIHNjLiJYdAMCKkXDduJ6U+9UoG4Ck3/WhmvYbAxL1KkYAiEAs6bW7GOVqwbseZOTJks0GkTavxp4QFxsOO2HYclqamY=
|