Overview
In one line, Orca is a local terminal coding Agent that treats DeepSeek as a first-class citizen: give it a task and it reads code, edits files, runs commands, and verifies results on its own until done or it needs your confirmation. Why try it DeepSeek-native: uses DeepSeek reasoning and tool-call semantics directly, SSE streaming, max reasoning strength by default, and prompt organization aimed at prefix caching (measured main-loop cache hit 99% after compression)—long tasks do not keep getting more expensive. 1M context self-management: 1M token window; auto-compresses past the 80% threshold while keeping the system prompt and recent turns so long-task context does not drop. Terminal-native experience: single Rust binary, no runtime deps. orca enters TUI interaction; orca exec "fix failing tests" runs scripts and CI directly; orca exec --verifier "cargo test" "fix it" includes a verification loop. Goal mode: /goal sets a long-term goal; after each success it continues automatically and can resume after restart; the model must pass goal audit before declaring done or blocked—suited to overnight tasks.
Connected profiles
More Coding assistants products
Sources · 0Sources & provenance
English text is maintained separately from the original record. Consult the original source before citing a fact.
No source links are available for this profile.
Original source text
Names may use phonetic renderings when no English name is supplied. The original wording is retained below.
一句话Orca 是一个把 DeepSeek 当作第一公民的本地终端编程 Agent：给它一个任务，它自己读代码、改文件、跑命令、验证结果，直到做完或者需要你确认。为什么值得一试DeepSeek 原生：直接使用 DeepSeek 的推理与工具调用语义，SSE 流式输出、默认最大推理强度、面向前缀缓存的提示词组织（实测压缩后主循环缓存命中 99%），长任务不会越跑越贵。1M 上下文自管理：1M token 窗口，超过 80% 阈值自动压缩，保留系统提示与最近轮次，长任务上下文不断档。终端原生体验：Rust 单二进制，无运行时依赖。orca 进入 TUI 交互；orca exec "修复失败的测试" 直接跑脚本与 CI；orca exec --verifier "cargo test" "修复它" 自带验证闭环。目标模式：/goal 设定长期目标，每轮成功后自动继续，重启后仍可恢复，模型必须经过目标审计才能宣布完成或阻塞，适合"跑一整晚"的任务。