@@ -17,7 +17,7 @@ A cross‑platform desktop UI (Avalonia/.NET 8) for driving the Codex CLI app se
1717## Requirements
1818
1919- .NET SDK 8.0+
20- - Codex CLI installed and on ` PATH `
20+ - Codex CLI v0.58.0 or newer installed and on ` PATH `
2121 - Verify with: ` codex app-server --help `
2222- No external Git required — uses LibGit2Sharp for repo init/staging/commit
2323
@@ -85,36 +85,36 @@ If you define profiles in `config.toml`, Semantic Developer surfaces them in the
8585Example ` config.toml ` profiles:
8686
8787``` toml
88- [profiles .gpt-5-high ]
89- model = " gpt-5"
88+ [profiles .gpt-5 . 1 -high ]
89+ model = " gpt-5.1 "
9090model_provider = " openai"
9191approval_policy = " never"
9292model_reasoning_effort = " high"
9393model_reasoning_summary = " auto"
9494
95- [profiles .gpt-5-medium ]
96- model = " gpt-5"
95+ [profiles .gpt-5 . 1 -medium ]
96+ model = " gpt-5.1 "
9797model_provider = " openai"
9898approval_policy = " never"
9999model_reasoning_effort = " medium"
100100model_reasoning_summary = " auto"
101101
102- [profiles .gpt-5-low ]
103- model = " gpt-5"
102+ [profiles .gpt-5 . 1 -low ]
103+ model = " gpt-5.1 "
104104model_provider = " openai"
105105approval_policy = " never"
106106model_reasoning_effort = " low"
107107model_reasoning_summary = " auto"
108108
109- [profiles .gpt-5-codex-high ]
110- model = " gpt-5-codex"
109+ [profiles .gpt-5 . 1 -codex-high ]
110+ model = " gpt-5.1 -codex"
111111model_provider = " openai"
112112approval_policy = " never"
113113model_reasoning_effort = " high"
114114model_reasoning_summary = " auto"
115115
116- [profiles .gpt-5-codex-medium ]
117- model = " gpt-5-codex"
116+ [profiles .gpt-5 . 1 -codex-medium ]
117+ model = " gpt-5.1 -codex"
118118model_provider = " openai"
119119approval_policy = " never"
120120model_reasoning_effort = " medium"
@@ -183,11 +183,13 @@ Notes
183183## Conversation & Protocol Behavior
184184
185185- Always uses the Codex app server: the app starts the CLI with ` codex app-server ` .
186- - User input is wrapped as a protocol ` Submission ` with a new ` id ` and an ` op ` payload:
187- - Defaults to ` user_input ` with ` items: [{ type: "text", text: "..." }] ` .
188- - When the app infers that a full turn is required, it sends ` user_turn ` and includes
189- ` cwd ` , ` approval_policy ` (defaults to ` on-request ` ), ` sandbox_policy ` (defaults to
190- ` workspace-write ` ), optional ` model ` , and default reasoning fields.
186+ - Session lifecycle uses the v0.58 thread/turn RPCs exclusively:
187+ - When a session begins (or resumes), the app issues ` thread/start ` / ` thread/resume `
188+ with ` approvalPolicy="onRequest" ` and ` sandbox="workspaceWrite" ` along with the
189+ selected ` model ` , ` cwd ` , and any reasoning overrides.
190+ - Each user message maps to a ` turn/start ` call that sends the active ` threadId ` ,
191+ a text ` input ` block, the desired reasoning effort, and a ` sandboxPolicy ` object
192+ (` { mode: "workspaceWrite", networkAccess: <bool> } ` ).
191193- Auto‑approval: when the CLI emits an ` exec_approval_request ` or ` apply_patch_approval_request ` ,
192194 the app automatically approves the request. There is no setting to toggle this.
193195
0 commit comments