refactor: remove TUI auto-spawn and integrate llm-agents codex

.github/actions/setup-nix/action.yaml

@@ -29,6 +29,6 @@ runs:
     - name: Setup nix develop env
       shell: bash
       run: |
-        echo 'use flake' >.envrc
+        echo 'use flake .#ci' >.envrc
     - uses: HatsuneMiku3939/direnv-action@v1
     # and avoid develop env being gc

crates/coco-tui/src/e2e_tests/support.rs

@@ -676,9 +676,6 @@ fn resolve_coco_bin_from_env() -> Option<PathBuf> {
     if let Ok(path) = env::var("COCO_TEST_BIN") {
         return Some(PathBuf::from(path));
     }
-    if let Ok(path) = env::var("COCO_TUI_BIN") {
-        return Some(PathBuf::from(path));
-    }
     if let Ok(path) = env::var("CARGO_BIN_EXE_coco") {
         return Some(PathBuf::from(path));
     }
@@ -703,7 +700,7 @@ pub(crate) fn coco_binary() -> PathBuf {
     let path = resolve_coco_bin_from_env().unwrap_or_else(resolve_coco_bin_from_target);
     assert!(
         path.exists(),
-        "coco binary not found at {:?}; build `cargo build -p coco-tui --bin coco` or set COCO_TUI_BIN/COCO_TEST_BIN",
+        "coco binary not found at {:?}; build `cargo build -p coco-tui --bin coco` or set COCO_TEST_BIN",
         path
     );
     path

crates/coco-tui/src/main.rs

@@ -134,11 +134,9 @@ impl TryFrom<Commands> for ClientCommand {
                 command,
             },
             Commands::Mcp { args } => ClientCommand::Mcp { args },
-            Commands::Combo(ComboCommands::Run { name, args }) => ClientCommand::ComboRun {
-                name,
-                args,
-                ignore_workspace_scripts: false,
-            },
+            Commands::Combo(ComboCommands::Run { name, args }) => {
+                ClientCommand::ComboRun { name, args }
+            }
         };
         Ok(command)
     }
@@ -175,14 +173,7 @@ async fn main() -> Result<()> {
 
         if should_handle_client {
             match ClientCommand::try_from(command) {
-                Ok(mut command) => {
-                    if let ClientCommand::ComboRun {
-                        ignore_workspace_scripts,
-                        ..
-                    } = &mut command
-                    {
-                        *ignore_workspace_scripts = args.ignore_workspace_scripts;
-                    }
+                Ok(command) => {
                     init_client_logging(&program, &command);
                     return handle_client_command(&program, &command_name, command)
                         .await

flake.nix

@@ -4,12 +4,17 @@
     nixpkgs.url = "github:NixOS/nixpkgs/nixpkgs-unstable";
     fenix.url = "github:nix-community/fenix";
     fenix.inputs.nixpkgs.follows = "nixpkgs";
+
+    jail-nix.url = "sourcehut:~alexdavid/jail.nix";
+    llm-agents.url = "github:numtide/llm-agents.nix";
   };
 
   outputs = {
     self,
     nixpkgs,
     utils,
+    jail-nix,
+    llm-agents,
     ...
   } @ inputs:
     utils.lib.eachDefaultSystem
@@ -104,18 +109,26 @@
           };
         };
         devShells = let
-          components = [
-            "cargo"
-            "clippy"
-            "rust-src"
-            "rustc"
-            "rustfmt"
-            "llvm-tools"
-            "rust-analyzer"
+          jail = jail-nix.lib.init pkgs;
+          commonPackages = with pkgs; [
+            bashInteractive
+            curl
+            wget
+            jq
+            git
+            which
+            ripgrep
+            gnugrep
+            gawkInteractive
+            ps
+            findutils
+            gzip
+            unzip
+            gnutar
+            diffutils
           ];
-          packages = with pkgs;
+          devPackages = with pkgs;
             [
-              # Development
               grcov
               prek
 
@@ -126,21 +139,53 @@
               run-test
               run-cov
             ]);
+          nativeBuildInputs = with pkgs; ([
+              (fenix.stable.withComponents [
+                "cargo"
+                "clippy"
+                "rust-src"
+                "rustc"
+                "rustfmt"
+                "llvm-tools"
+                "rust-analyzer"
+              ])
+            ]
+            ++ lib.optionals stdenv.isLinux [pkg-config]);
+          commonJailOptions = with jail.combinators; [
+            network
+            time-zone
+            no-new-session
+            mount-cwd
+          ];
+          codex = llm-agents.packages.${system}.codex;
+          jailed-codex = jail "jailed-codex" codex (with jail.combinators; (
+            commonJailOptions
+            ++ [
+              (readwrite (noescape "~/.codex"))
+              (add-pkg-deps [pkgs.stdenv.cc])
+              (add-pkg-deps commonPackages)
+              (add-pkg-deps nativeBuildInputs)
+              (add-pkg-deps devPackages)
+            ]
+          ));
+          packages = commonPackages ++ devPackages ++ lib.optionals pkgs.stdenv.isLinux [jailed-codex];
+          shellHook = ''
+            # Unset SOURCE_DATE_EPOCH to prevent reproducible build timestamps during development.
+            # This allows timestamps to reflect the current time, which is useful for development workflows.
+            unset SOURCE_DATE_EPOCH
+          '';
         in rec {
           default = stable;
-          stable = pkgs.mkShell {
-            nativeBuildInputs = with pkgs; ([
-                (fenix.stable.withComponents components)
-              ]
-              ++ lib.optionals stdenv.isLinux [pkg-config]);
+          ci = pkgs.mkShell {
+            inherit nativeBuildInputs;
+            inherit shellHook;
 
+            packages = devPackages;
+          };
+          stable = pkgs.mkShell {
+            inherit nativeBuildInputs;
             inherit packages;
-
-            shellHook = ''
-              # Unset SOURCE_DATE_EPOCH to prevent reproducible build timestamps during development.
-              # This allows timestamps to reflect the current time, which is useful for development workflows.
-              unset SOURCE_DATE_EPOCH
-            '';
+            inherit shellHook;
           };
         };
       }

src/cli.rs

@@ -22,7 +22,6 @@ pub enum ClientCommand {
     ComboRun {
         name: String,
         args: Vec<String>,
-        ignore_workspace_scripts: bool,
     },
     Record {
         wrap_result: bool,
@@ -69,11 +68,7 @@ pub async fn handle_client_command(
         ClientCommand::Reply { fields } => crate::cmd::handle_reply(fields)
             .await
             .whatever_context("failed to handle reply"),
-        ClientCommand::ComboRun {
-            name,
-            args,
-            ignore_workspace_scripts,
-        } => crate::cmd::handle_combo_run(name, args, ignore_workspace_scripts)
+        ClientCommand::ComboRun { name, args } => crate::cmd::handle_combo_run(name, args)
             .await
             .whatever_context("failed to handle combo run"),
         ClientCommand::Record {