v0.16.1

Features

• feat(context-enhancement): correct and update to improve context enhanced message
• feat(tools): re-work on read and write file tools for more tolerance to small model
• feat(tools): every tokens counted, compact the tool result to reduce token consumed
• feat(tools): rename file tools for better usability from llm
• feat(llm): make tool parsing more tolerance for small models
• feat(memory): restricted memory more
• feat(files): use relative path for file-tools to reduce token consumption

Fixes

• fix(agent-plan): fixed issue that agent plan can be triggered in the middle of message
• fix(llm): setup model from last provider
• fix(job): correct how job message is generated

Chores

• chore(context): hided absolute path information from context
• chore(command_execution): /dev/null redirect should not be blocked
• chore(file_handler): defer the docling loading only if we need them to increase load speed
• chore(editing): guardrail the file edit to prevent malform edit
• chore(custom_llm): fixed tool and service to harness for small model
• chore(stream-session): increase the timeout for waiting stream
• chore(llm): add crofai to provider list
• chore(models): update models list for some provider
• chore(get_file): increase the line length limit
• chore(llm): add fast model for dependent services