100 Commits

This Branch

This Branch

All Branches

Author	SHA1	Message	Date
Dark-Alex-17	35e75e5b4f	fix: ctrl-c inside of an auto-continue loop created an infinite loop	2026-07-31 11:44:16 -06:00
Dark-Alex-17	4f38214681	fix: sbx update doesn't allow undefined fields in sbx spec	2026-07-31 10:07:29 -06:00
Dark-Alex-17	69e1b98c44	fix: ctrl-c interruption doesn't discard session messages when throbber is showing	2026-07-30 13:38:37 -06:00
Dark-Alex-17	0324436114	feat: ctrl-c interrupts ongoing prompt in a session, but lets the user inject more instructions mid-stream	2026-07-30 10:21:19 -06:00
Dark-Alex-17	233c212d2a	feat: improved function calling performance by allowing parallel tool calling	2026-07-29 12:40:52 -06:00
Dark-Alex-17	e288b41365	fix: improper handling of fd-style globbing for directories in fs_glob	2026-07-29 12:40:33 -06:00
Dark-Alex-17	fbf6a6bdf4	fix: properly templated architect design doc path in starter commands	2026-07-29 11:59:09 -06:00
Dark-Alex-17	57b72702b2	feat: created the architect and gatekeeper agents for dramatically improved coding performance	2026-07-29 11:23:52 -06:00
Dark-Alex-17	38ba303c3c	feat: Improved readability of session message exchange replays	2026-07-29 10:15:35 -06:00
Dark-Alex-17	fc7bc0ff8f	fix: .copy works when sessions are resumed	2026-07-29 09:50:40 -06:00
Dark-Alex-17	7f7ea758a7	fix: ACP session/prompt now drives the full tool-execution loop ... run_prompt_turn previously performed a single completion and returned, leaving tool calls unexecuted. It now mirrors start_directive's loop: call -> execute tools -> merge results -> continue until no tool results, then check the pending-agents guardrail before returning. The pending-agents guardrail injects a reminder prompt when sub-agents are pending, matching the same loop termination semantics as --headless. The session is NOT exited between turns (unlike start_directive) so multi-turn ACP conversations retain history across session/prompt calls. Manual gate (tool-probe agent, fs_write + fs_read tools): id 3 result: {"output":"DONE:probe.txt","stopReason":"end_turn"} probe.txt exists: YES, content: hello	2026-07-28 15:42:51 -06:00
Dark-Alex-17	06b2c384e3	fix: ACP spec conformance — ContentBlock prompt params and protocolVersion type ... DEFECT 1: session/prompt now accepts the spec-shaped params as primary: {"prompt": [{"type": "text", "text": "..."}]} Text blocks are joined with newlines. Non-text block types are silently ignored. The legacy params.text alias is preserved as a fallback. -32602 is returned only when neither a non-empty prompt array with text blocks nor a non-empty text field is present. DEFECT 2: initialize result now emits protocolVersion as integer 1 instead of the string "1", matching the ACP spec's InitializeResponse. Tests: 4 new unit tests pin the spec-shaped prompt path, the non-text block ignore behavior, the missing-both -32602 path, and the numeric protocolVersion type.	2026-07-28 14:41:22 -06:00
Dark-Alex-17	d50de7c06a	lint: fixed test ordering	2026-07-27 19:52:51 -06:00
Dark-Alex-17	0956f08791	refactor: move ACP server dispatch into run() for shared flag setup ... All CLI flag processing (--agent, --role, --rag, --model, --no-memory, --no-stream, --no-workspace-instructions, etc.) now runs through run() before the REPL/cmd split. The ACP server is dispatched right before match is_repl, after apply_prelude and skills loading, so it benefits from the complete context setup with no duplication or drift risk.	2026-07-27 19:44:11 -06:00
Dark-Alex-17	087d0c320c	feat: apply --agent/--role/--rag/--model flags in --acp-server mode ... These CLI flags were previously ignored because the ACP branch returned before run() could apply them. Now the context is configured with the requested agent, role, RAG index, and model before the server starts. Session management remains protocol-driven via session/new and session/load.	2026-07-27 19:31:45 -06:00
Dark-Alex-17	2128390f99	fmt: applied formatting	2026-07-27 19:26:29 -06:00
Dark-Alex-17	d008de1848	fix: restore stdout output for standalone --headless mode ... RenderMode::Silent was incorrectly applied to --headless in addition to --acp-server. Standalone headless should still display the LLM response; only ACP mode requires stdout purity for JSON-RPC. The acp server's run_prompt_turn already sets Silent before each prompt call.	2026-07-27 19:22:51 -06:00
Dark-Alex-17	d79787bf96	fix: suppress tool-call display in headless mode; initialize session on session/new	2026-07-27 19:14:28 -06:00
Dark-Alex-17	577c51b62f	fix: skip stdin drain and set silent render mode when --acp-server is active	2026-07-27 18:59:09 -06:00
Dark-Alex-17	d462b09f80	feat: add headless profile to sbx-kit spec	2026-07-27 18:03:35 -06:00
Dark-Alex-17	b711e4983b	feat: implement ACP user-interaction to request_permission bridge	2026-07-27 18:00:07 -06:00
Dark-Alex-17	6ae3efb06c	feat: implement ACP session/load and session/cancel	2026-07-27 17:52:49 -06:00
Dark-Alex-17	02dd14394b	feat: implement ACP session/prompt	2026-07-27 17:48:31 -06:00
Dark-Alex-17	f11d4ca760	feat: add ACP server skeleton with stdout-purity test	2026-07-27 17:31:47 -06:00
Dark-Alex-17	f1415067f2	feat: add --headless flag for unattended operation	2026-07-27 17:17:56 -06:00
Dark-Alex-17	af5c34fde5	Merge branch 'main' of github.com:Dark-Alex-17/coyote	2026-07-27 15:05:25 -06:00
Dark-Alex-17	2ffa278f2d	test: testing potential nerdbox regression fix for coyote sandbox mode	2026-07-27 15:04:41 -06:00
Dark-Alex-17	f4cbee9611	docs: remove comment in spec.yaml about copying in coyote password file	2026-07-27 10:50:40 -06:00
Dark-Alex-17	e95fd1e06e	ci: fix typo in coyote image tag; needs leading 'v'	2026-07-27 09:51:23 -06:00
Dark-Alex-17	d79ea55e09	fix: include graph-agent descriptions in .agent <TAB> completions	2026-07-27 09:34:36 -06:00
Dark-Alex-17	58d9d4c64e	docs: updated help message for --fresh flag	2026-07-24 18:30:45 -06:00
Dark-Alex-17	63a768aa46	fix: fresh wizard openai-compatible support	2026-07-24 18:29:24 -06:00
Dark-Alex-17	2c3f671efa	fix: config existence check for --fresh sandboxes	2026-07-24 18:09:06 -06:00
Dark-Alex-17	8b0a536f4e	feat: Dynamically detect if a selected client in the sandbox first run wizard supports oauth	2026-07-24 18:07:30 -06:00
Dark-Alex-17	5a1bb569b4	feat: Add support for the --fresh flag again with host environment configuration injection	2026-07-24 18:00:41 -06:00
Dark-Alex-17	0c12580836	fix: Improve coyote sandbox startup time	2026-07-24 17:23:08 -06:00
Dark-Alex-17	df909325a7	fmt: applied formatting	2026-07-24 16:50:27 -06:00
Dark-Alex-17	d379fcddf8	fix: bypass forgotten sandbox mode check for MCP secret interpolation	2026-07-24 16:50:04 -06:00
Dark-Alex-17	b405acd8f0	feat: force overwrite global sbx secrets	2026-07-24 16:45:51 -06:00
Dark-Alex-17	805ae7112a	feat: add sbx secrets globally	2026-07-24 16:41:00 -06:00
Dark-Alex-17	6ddbf37523	feat: only create secrets local to a sandbox	2026-07-24 16:36:16 -06:00
Dark-Alex-17	f510bb649b	feat: Improved credentials management for docker sandboxes	2026-07-24 16:26:35 -06:00
Dark-Alex-17	c1b14bdfdf	docs: Created a mermaid diagram for Sisyphus	2026-07-24 13:55:31 -06:00
Dark-Alex-17	46f2a9eae2	docs: added in forgotten connection to the librarian agent diagram	2026-07-24 13:44:07 -06:00
Dark-Alex-17	cf3a12141b	docs: updated the graph agent diagrams to use mermaid diagrams for more easily readable diagramming in their READMEs	2026-07-24 13:39:59 -06:00
Dark-Alex-17	8f13810f0f	chore: updated models.yaml	2026-07-24 13:27:20 -06:00
Dark-Alex-17	eba8c86e21	fix: properly wrap sub-style changes in markdown rendering	2026-07-24 12:13:33 -06:00
Dark-Alex-17	d75fb47de1	docs: Updated project license	2026-07-23 21:07:51 -06:00
Dark-Alex-17	49a281bf93	feat: renamed --contents arg for fs_write/patch to --content since most models attempt that first and error otherwise	2026-07-23 19:13:36 -06:00
Dark-Alex-17	b9474ce6ef	feat: Used improved theme selections for tool call highlighting colors	2026-07-23 18:51:57 -06:00
Dark-Alex-17	d328310880	feat: Improved theme-derived syntax highlighting for LLM tool call logging	2026-07-23 18:45:23 -06:00
Dark-Alex-17	5eb77ab467	feat: Improved coloring of LLM tool invocation outputs to make LLM output more readable and cohesive	2026-07-23 18:27:54 -06:00
Dark-Alex-17	a2c4f05c8c	feat: renamed the user__ask to user__select and improved descriptions to improve model usage	2026-07-23 18:22:18 -06:00
Dark-Alex-17	89fadcca15	feat: Improved coloring/highlighting of tool calls to make LLM invocation logs easier to read	2026-07-23 18:12:18 -06:00
Dark-Alex-17	ab3a818507	docs: Added the new compression control fields to the config example file	2026-07-23 12:40:38 -06:00
Dark-Alex-17	63f73f22c3	style: applied formatting to new long run features	2026-07-23 12:30:30 -06:00
Dark-Alex-17	54c5079cb7	feat: long-running session improvements ... - Hang fix: 120s timeout on compression LLM call + raw_stream channel-close handling (None => break instead of spinning) - Supervisor effective active count: stop counting finished JoinHandles as occupying capacity slots - Universal tool result size cap: truncate_if_needed() on ToolResult, applied in eval_tool_calls() after escalation block; configurable via max_tool_result_chars in AppConfig/AgentConfig - Windowed compression: compression_keep_last config param keeps the N most recent messages visible after compression - Fix pre-existing flaky test: add #[serial] to handle_list_available_unrestricted_when_no_whitelist so it does not race with TestConfigDirGuard-based tests that temporarily populate the agents data dir	2026-07-23 12:25:54 -06:00
Dark-Alex-17	d51bdd3086	feat: Made sisyphus suite of agents all auto-approved for tool usage	2026-07-22 22:28:20 -06:00
Dark-Alex-17	56ec58a748	fix: removed accidental duplicate ast_grep tool in explore agent	2026-07-22 17:42:14 -06:00
Dark-Alex-17	93f9c5425e	fix: npm and npx need the /usr/local/share/npm-global/lib directory to exist to run properly so I've added it to the dockerfile	2026-07-22 17:11:57 -06:00
Dark-Alex-17	77dfd08103	fix: added executable bit to adversary agent tools script	2026-07-22 16:38:42 -06:00
Dark-Alex-17	46dcef0dec	Merge branch 'main' of github.com:Dark-Alex-17/coyote	2026-07-22 16:37:34 -06:00
Dark-Alex-17	72c6bb74c2	feat: created the adversay agent and adversarial-review skill	2026-07-22 16:35:00 -06:00
Dark-Alex-17	6bf80dcce9	feat: added ast_grep tool to Sisyphus suite agents	2026-07-22 15:41:09 -06:00
Dark-Alex-17	df948c69bf	feat: new spawnable_agents field in agents to let users restrict what agents can be spawned by a parent agent	2026-07-22 14:29:55 -06:00
Dark-Alex-17	e5d0fcc764	test: fixed linter issues on markdown tests	2026-07-22 13:32:37 -06:00
Dark-Alex-17	c60d3e7cda	style: updated some stylistic things across the new markdown rendering implementation	2026-07-22 13:20:56 -06:00
Dark-Alex-17	c21fd47b42	feat: replay pre-compressed messages as well when resuming sessions for users to see	2026-07-22 13:05:06 -06:00
Dark-Alex-17	ab0b89dd67	fix: fetch descriptions from graph agent configs as well when listing agents	2026-07-22 12:59:30 -06:00
Dark-Alex-17	420db4bb88	feat: created a new builtin function for agents who can spawn other agents to list available agents via agent__list_available	2026-07-22 12:50:41 -06:00
Dark-Alex-17	dfacf31f6a	fix(render): suppress blank lines above rendered table	2026-07-22 12:37:30 -06:00
Dark-Alex-17	8cfd5ee2c4	docs(render): update phase 2.7 SHA after amend	2026-07-22 12:30:34 -06:00
Dark-Alex-17	e82e5ab8e4	test(render): comprehensive table rendering coverage	2026-07-22 12:30:10 -06:00
Dark-Alex-17	d790782ace	feat(render): hanging-indent line wrapping for lists and blockquotes	2026-07-22 12:27:13 -06:00
Dark-Alex-17	bf06d5e8f3	feat(render): wire table state machine and finalize hook	2026-07-22 12:23:44 -06:00
Dark-Alex-17	cdfaa0f111	feat(render): render markdown tables with comfy-table	2026-07-22 12:20:02 -06:00
Dark-Alex-17	c062f34852	feat(render): parse table cells and column alignments	2026-07-22 12:16:05 -06:00
Dark-Alex-17	7671d28d6e	feat(render): detect markdown table rows and separators	2026-07-22 12:14:45 -06:00
Dark-Alex-17	fcc4a1d2b5	feat(render): add comfy-table dependency and table border style	2026-07-22 12:11:04 -06:00
Dark-Alex-17	b0eeba110d	test(render): comprehensive coverage for rich markdown renderer	2026-07-22 11:48:43 -06:00
Dark-Alex-17	d65d63ee50	feat(render): activate rich markdown renderer as default	2026-07-22 11:47:39 -06:00
Dark-Alex-17	9890cf0ddc	feat(render): rich block-level markdown rendering (headings, quotes, lists, hr)	2026-07-22 11:45:03 -06:00
Dark-Alex-17	89db5b3887	feat(render): rich inline markdown rendering (bold, italic, code, links)	2026-07-22 11:41:35 -06:00
Dark-Alex-17	f40ba4ccbe	feat(render): detect markdown block-level line types	2026-07-22 11:37:11 -06:00
Dark-Alex-17	d2940a8d32	feat(render): precompute markdown scope styles for rich rendering	2026-07-22 11:31:23 -06:00
Dark-Alex-17	ed7ad36475	feat: Created the raw_markdown configuration flag	2026-07-22 11:03:27 -06:00
Dark-Alex-17	393ed16963	fix: chown the full sandbox cache dir, not just the coyote subdir	2026-07-22 10:08:00 -06:00
Dark-Alex-17	3f94d2003a	fix: chown the whole coyote cache dir not just the oauth dir in the sandbox	2026-07-22 09:48:11 -06:00
Dark-Alex-17	50ff9008fe	chore: updated deepseek models	2026-07-22 09:31:17 -06:00
Dark-Alex-17	9606d7f8aa	feat: added a .fork command to fork a new session from a running conversation	2026-07-21 14:54:55 -06:00
Dark-Alex-17	d8ae9e25d9	docs: reworded explore agent instructions to empower agent to spawn as many sub agents as it deems necessary	2026-07-21 14:14:00 -06:00
Dark-Alex-17	006f64bfa0	docs: Improved sisyphus wording to empower agent to spawn as many subagents as necessary	2026-07-21 14:12:51 -06:00
Dark-Alex-17	000559bc9d	style: Applied formatting	2026-07-21 11:24:18 -06:00
Dark-Alex-17	3aede58a11	feat(oauth): enable browser-paste PKCE flow for OpenAI-compatible providers ... Two coordinated changes that make openai-compatible OAuth providers usable with a non-localhost redirect_uri (browser shows the callback URL, user copies it back into the terminal — the same UX Claude uses). Fix: OpenAICompatibleOAuthProvider::fixed_redirect_uri() previously returned Some(uri) for any redirect_uri including public HTTPS URLs, which trapped run_pkce_flow into trying to bind a TCP listener on a public URL. It now returns Some only for loopback URIs (127.0.0.1, localhost, ::1). Non-loopback URIs return None, routing run_pkce_flow to the paste branch. New tri-format paste parser (parse_paste_input): - Full callback URL (starts with http:// or https://): parse code + state from the query string. This is what most modern OAuth providers redirect to and what a naive user copies from the browser bar. - Anthropic-style code#state fragment: preserved for Claude compatibility. - Bare code: accepted with a warning that CSRF state validation is skipped. For providers whose callback page shows only the code with no state. State validation moved from mandatory to conditional — if a paste didn't carry state (bare-code path), we warn and skip the check instead of hard- failing. The listener path (localhost + LAN redirects) still requires state because the server sends it in the query. Adds 9 unit tests covering both changes.	2026-07-21 11:14:55 -06:00
Dark-Alex-17	cab1e72b97	fix: fix typo in Gemini's generation_config property to use camelCase exclusively	2026-07-21 11:11:00 -06:00
Dark-Alex-17	cd4bf245e9	chore: updated models.yaml	2026-07-21 11:09:51 -06:00
Dark-Alex-17	82bf6176f8	fix(oauth): treat missing expires_in as non-expiring device_code token ... GitHub OAuth Apps issue tokens that never expire and omit expires_in from the response (they only send access_token, token_type, scope). RFC 6749 §5.1 allows this — expires_in is only REQUIRED for tokens that actually expire. When expires_in is missing, save the token with expires_at = i64::MAX so prepare_oauth_access_token never tries to refresh. If the token is ever revoked server-side, the eventual 401 on the API call is the user's cue to re-authenticate. No effect on providers that include expires_in (Moonshot etc. — unchanged).	2026-07-21 10:32:47 -06:00
Dark-Alex-17	d407eb5a6a	fix(oauth): send Accept: application/json in device flow requests ... GitHub's device flow endpoints (and likely other RFC 8628 servers) default to responding in application/x-www-form-urlencoded unless the client asks for JSON via the Accept header. Our device auth and polling paths both call .json() on the response and were failing to decode form-urlencoded bodies with 'expected value at line 1 column 1'. Adds Accept: application/json to: - The device authorization POST in run_device_code_flow - The device_code polling POST (on the RequestBuilder returned by build_token_request) RFC 6749 §5.1 already specifies JSON as the token response format, so this is spec-compliant across providers. Servers that already default to JSON (Moonshot, etc.) ignore the redundant header.	2026-07-21 10:30:36 -06:00
Dark-Alex-17	6f2594712f	refactor: Standardized paths module function names to not use 'path' in the name and to just always be either 'dir' or 'file'	2026-07-21 10:15:42 -06:00
Dark-Alex-17	79d43c8791	docs: config.example.yaml example for Device Authorization Grant ... Adds a commented example under the openai-compatible client section showing how to configure flow: device_code for RFC 8628 device flow. Uses Moonshot's kimi-code endpoints as the illustrative reference (users supply their own client_id — no bundled defaults per §5.8 of the design plan).	2026-07-20 15:27:20 -06:00