v0.0.7

New Features

• Move prompt envelope, sigil refs, variables, and #ai static media into aidialog core; bump context image budget to 768px (#17)
• Resize images for LLM context and pick one image per output, via shared output_parts/merge_media helpers (#13)
• Move join_out/output_from_msg to fastcore.nbio, add render_md output-to-markdown renderer, and add %nbrun line magic (#12)
• Support nbdev meta directives: export= in add_msg/mk_message, [export] in previews, and directive attrs in XML views (#11)
• url_mime: use server Content-Type as fallback before byte sniffing (#10)

v0.0.6

New Features

• Add SVG mime type and ensure url_mime falls back to default when sniffing fails (#9)

v0.0.5

New Features

• Add hist module with dialog↔chat history conversions, migrated from llmsurgery (#7)

Bugs Squashed

• Add ToolResponse, moved from fastllm (#8), thanks to @curtis-allan

v0.0.4

New Features

• Show a prompt reply on its own preview line, with size hints for truncated content (#6)
• Split export state into a read-only merged exported and a host-owned meta_exported switch (#5), thanks to @jph00

v0.0.3

New Features

• Add the msg_parts message model and msg2md/dlg2md Markdown rendering (#4), thanks to @jph00
• Add folded-display rendering for tool calls and token usage blocks (#3)

v0.0.2

New Features

• Move .hist and .compact back to llmsurgery, consolidating other functions into .dialog and replacing nbformat with fastcore.nbio (#2)

v0.0.1

New Features

• Migrate from llmsurgery (#1)