| 文件 | 最后提交记录 | 最后更新时间 |
|---|---|---|
feat(sandbox): introduce /allow-path command for granting sandbox path permissions - Added a new TUI command /allow-path to allow users to grant read or write permissions for specific paths in the local sandbox. - Implemented functionality to handle path resolution and permission granting within the GatewayRuntime. - Enhanced the permission control interface to support sandbox denial checks for specified paths. - Updated documentation to reflect the new command and its usage. This feature improves user control over sandbox permissions, allowing for more flexible interaction with the local environment. | 21 天前 | |
optimize: add proactive token estimation and config validation with URL optimization Co-Authored-By: GLM-5 | 16 天前 | |
feat(llm-client): thread cached input tokens through Usage into trace spans Account for cache-read tokens end to end: - Usage carries cached_tokens (OpenAI prompt_tokens_details.cached_tokens; Anthropic cache_read_input_tokens), threaded into LLM_CALL / TURN trace spans. - Anthropic marks cache_control ephemeral on the stable system-prefix block ONLY, never the volatile Context block, so the breakpoint is read back across turns instead of being rewritten every turn. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com> | 12 天前 | |
fix(cerberus): decode network event port in host byte order The eBPF program already converts sin_port from network to host byte order ( u16::from_be) before writing it into the perf event buffer, so the user-space decoder must read it back in native order. It was instead re-applying a big-endian conversion (u16::from_be_bytes), byte-swapping the port a second time (e.g. 443 surfaced as 47873). Read the field with u16::from_ne_bytes, consistent with the other fields decoded from the same buffer. This had no visible effect until the network audit surfaced the port to users via NetworkAccessRecord. | 20 天前 | |
chore: remove end-to-end report and script for generating it | 2 个月前 | |
feat: enhance external command handling in TUI and add tests - Refactored external command processing to improve clarity and functionality. - Introduced new functions for expanding commands and handling user arguments. - Added unit tests to ensure correct behavior of external command expansion. - Updated Cargo.lock to include new dependencies: libc and tracing. | 9 天前 | |
feat(llm-client): thread cached input tokens through Usage into trace spans Account for cache-read tokens end to end: - Usage carries cached_tokens (OpenAI prompt_tokens_details.cached_tokens; Anthropic cache_read_input_tokens), threaded into LLM_CALL / TURN trace spans. - Anthropic marks cache_control ephemeral on the stable system-prefix block ONLY, never the volatile Context block, so the breakpoint is read back across turns instead of being rewritten every turn. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com> | 12 天前 | |
docs: update README and CLI documentation for reasoning effort feature - Added detailed explanation of the reasoning_effort configuration in the README. - Updated CLI documentation to include --reasoning-effort argument for provider-side reasoning levels. - Enhanced TUI interface to display current reasoning effort level and allow switching between levels. | 1 个月前 | |
feat: implement per-command process groups with registry for reliable cleanup Co-Authored-By: GLM-5 | 10 天前 | |
Make operation backend local-only | 1 个月前 | |
add rustls-tls-native-roots | 27 天前 | |
feat: implement per-command process groups with registry for reliable cleanup Co-Authored-By: GLM-5 | 10 天前 | |
fix(prompt): merge stable+volatile into one system message; sort skills deterministically Two reviewer findings addressed: - builder_impl: fold volatile tail into the single system message (stable first, volatile appended with \n\n) instead of emitting two separate system blocks. OpenAI-compatible providers including OpenRouter often expect one leading system message; multiple consecutive system roles can be silently dropped or mis-parsed by chat-template backends. Auto-prefix caching is unaffected: the stable span remains a byte-identical leading prefix every turn. - loader: sort loaded skills by name before returning, matching the tools.sort_by precedent in builder_impl. Unsorted read_dir output is FS/inode-order-dependent, so the skill catalog block in the stable prefix could differ between restarts or machines and defeat caching even when no skill changed. Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com> | 11 天前 | |
fix(prompt): merge stable+volatile into one system message; sort skills deterministically Two reviewer findings addressed: - builder_impl: fold volatile tail into the single system message (stable first, volatile appended with \n\n) instead of emitting two separate system blocks. OpenAI-compatible providers including OpenRouter often expect one leading system message; multiple consecutive system roles can be silently dropped or mis-parsed by chat-template backends. Auto-prefix caching is unaffected: the stable span remains a byte-identical leading prefix every turn. - loader: sort loaded skills by name before returning, matching the tools.sort_by precedent in builder_impl. Unsorted read_dir output is FS/inode-order-dependent, so the skill catalog block in the stable prefix could differ between restarts or machines and defeat caching even when no skill changed. Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com> | 11 天前 | |
feat(subagent): slim read-only subagent profile, turn budget, static repo map - Subagents get a dedicated slim system prompt and are restricted to read-only search tools (file_read/glob/grep), enforced even under an explicit --tools allowlist; skills are withheld from the child. - Caller-settable turn budget with a larger exploration default (~20); a fan-out prompt issues independent grep/glob/file_read in one turn. - A bounded static repository map is injected once into the MAIN agent's cache-stable prefix; subagents are excluded. - The subagent prompt gains a "disposable exploration worker" distillation instruction; the coordinator prompt-builder tests move to substring checks. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com> | 12 天前 | |
| 10 天前 | ||
fix(tui): input auto-wrap, Alt+Enter newline, and intelligent arrow key navigation By the way, apply cargo fmt for code style consistency. Co-Authored-By: GLM-5 | 13 天前 |
| 文件 | 最后提交记录 | 最后更新时间 |
|---|---|---|
| 21 天前 | ||
| 16 天前 | ||
| 12 天前 | ||
| 20 天前 | ||
| 2 个月前 | ||
| 9 天前 | ||
| 12 天前 | ||
| 1 个月前 | ||
| 10 天前 | ||
| 1 个月前 | ||
| 27 天前 | ||
| 10 天前 | ||
| 11 天前 | ||
| 11 天前 | ||
| 12 天前 | ||
| 10 天前 | ||
| 13 天前 |