feat: rename claude_model to task_model, address unmerged #272 review findings by umputun · Pull Request #284 · umputun/ralphex

umputun · 2026-04-16T22:46:51Z

Picks up unmerged follow-up commits from feat/claude-model-config (those 3 commits ended up on the branch but didn't make it into the squash-merge of #272).

Changes:

rename claude_model -> task_model and --claude-model -> --task-model for symmetry with review_model / --review-model
pass task model to plan mode (was missing in original PR)
strip duplicate --model flags from claude_args to avoid CLI errors when users put --model in args
fix typo in pkg/config/defaults/config: review_model section had # claude_model = example
add processor test for ReviewClaude executor separation (nil fallback + distinct review executor)
add ReviewModel merge test, normalize --flag=value form in docs
update README, CLAUDE.md, llms.txt, custom-providers.md
update review-claude tests for the post-codex-skip behavior added in Skip post-codex claude review when no findings #282

The claude_model name was introduced by the partial merge of #272, so the rename is correcting an in-flight feature rather than a stable API.

… findings - rename claude_model -> task_model and --claude-model -> --task-model for symmetry with review_model/--review-model - pass task model to plan mode (was missing) - strip duplicate --model flags from claude_args to prevent CLI errors - fix typo in pkg/config/defaults/config: review_model section had `# claude_model =` example - add processor test for ReviewClaude executor separation (nil fallback + distinct review executor) - add ReviewModel merge test, normalize --flag=value form in docs - update docs (README, CLAUDE.md, llms.txt, custom-providers.md) - update review-claude tests for post-codex-skip behavior added in #282 picks up unmerged follow-up commits from feat/claude-model-config branch

cloudflare-workers-and-pages · 2026-04-16T22:49:35Z

Deploying ralphex with Cloudflare Pages

Latest commit:	`5076249`
Status:	⚡️ Build in progress...

View logs

Copilot

Pull request overview

Renames the per-phase Claude model configuration from claude_model to task_model, wires the task model into plan mode, and hardens model flag injection to avoid duplicate --model CLI arguments.

Changes:

Rename config/CLI surface area to task_model / --task-model for symmetry with review_model / --review-model.
Ensure plan mode uses the resolved task model and add tests validating ReviewClaude executor routing/fallback.
Strip pre-existing --model entries from claude_args before injecting the configured model, and update docs to a consistent --flag=value style.

Reviewed changes

Copilot reviewed 13 out of 13 changed files in this pull request and generated 6 comments.

Show a summary per file

File	Description
pkg/processor/runner.go	Renames config fields and ensures task vs review executors use TaskModel/ReviewModel correctly.
pkg/processor/runner_test.go	Adds coverage for review executor separation and nil fallback behavior.
pkg/executor/executor.go	Adds `stripFlag` and uses it to avoid duplicate `--model` injection.
pkg/executor/executor_test.go	Adds tests for `stripFlag` and model flag override behavior.
pkg/config/values.go	Renames parsed key to `task_model` and merges TaskModel.
pkg/config/values_test.go	Updates parsing/merge tests for `task_model` and expands ReviewModel merge tests.
pkg/config/config.go	Renames exported config JSON key to `task_model` and wires through loader.
pkg/config/defaults/config	Updates default config template comments/examples for `task_model` and `review_model`.
cmd/ralphex/main.go	Renames CLI flag to `--task-model` and passes resolved task model to runner + plan mode.
README.md	Updates examples/options table to include `--task-model` and normalize `--flag=value` usage.
CLAUDE.md	Updates project conventions/docs examples for new flags and `--flag=value` style.
llms.txt	Updates usage examples and adds per-phase model configuration docs snippet.
docs/custom-providers.md	Documents optional `--model` injection in the ClaudeExecutor command line.

Comments suppressed due to low confidence (1)

pkg/config/values.go:189

Config parsing now only reads "task_model". Since the immediately-previous key was "claude_model", existing configs on the current mainline/branch will silently stop applying the model override after this rename. Consider accepting both keys (e.g., read task_model first, then fall back to claude_model) and optionally emitting a deprecation warning when claude_model is used.

	// claude settings
	if key, err := section.GetKey("claude_command"); err == nil {
		values.ClaudeCommand = key.String()
	}
	if key, err := section.GetKey("claude_args"); err == nil {
		values.ClaudeArgs = key.String()
	}
	if key, err := section.GetKey("task_model"); err == nil {
		values.TaskModel = key.String()
	}
	if key, err := section.GetKey("review_model"); err == nil {
		values.ReviewModel = key.String()
	}

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

Copilot · 2026-04-16T22:51:09Z

- `--wait` flag enables rate limit retry with specified duration (e.g., `--wait 1h`)
- `--session-timeout` flag sets per-session timeout for claude (e.g., `--session-timeout 30m`), kills hanging sessions
- `--idle-timeout` flag kills claude sessions when no output is received for a specified duration (e.g., `--idle-timeout 5m`), resets on each output line
+- `--task-model` flag sets model for task execution (e.g., `--task-model=opus`); `--review-model` sets model for review phases (falls back to `--task-model`). Only applies to default Claude CLI, ignored by custom wrappers (`claude_command`)


The docs here say task_model/review_model are ignored by custom wrappers, but ralphex will still pass a --model flag when these are set; wrappers may ignore it or translate it to their own model selection. Consider adjusting wording to avoid implying it can never work with wrappers, and (if desired) document that provided wrappers currently ignore it unless extended.

Suggested change

- `--task-model` flag sets model for task execution (e.g., `--task-model=opus`); `--review-model` sets model for review phases (falls back to `--task-model`). Only applies to default Claude CLI, ignored by custom wrappers (`claude_command`)

- `--task-model` flag sets model for task execution (e.g., `--task-model=opus`); `--review-model` sets model for review phases (falls back to `--task-model`). ralphex passes these through as `--model`; with custom wrappers (`claude_command`), support depends on the wrapper, which may ignore the flag or translate it to its own model-selection mechanism. The provided wrappers currently ignore it unless extended

Copilot · 2026-04-16T22:51:09Z

+| `task_model` | Model for task execution (e.g., `opus`, `sonnet`). Only applies to default Claude CLI, ignored by custom wrappers | empty |
+| `review_model` | Model for review phases. Falls back to `task_model` if empty. Only applies to default Claude CLI | empty |


The README claims task_model/review_model are ignored by custom wrappers, but the executor will pass --model whenever these are set and wrapper behavior determines whether it is ignored or mapped. Consider clarifying that this is wrapper-dependent (and that wrappers should accept/ignore unknown flags to avoid breaking when --model is present).

Suggested change

| `task_model` | Model for task execution (e.g., `opus`, `sonnet`). Only applies to default Claude CLI, ignored by custom wrappers | empty |

| `review_model` | Model for review phases. Falls back to `task_model` if empty. Only applies to default Claude CLI | empty |

| `task_model` | Model for task execution (e.g., `opus`, `sonnet`). Intended for the default Claude CLI; when set, the executor may still pass `--model` to custom wrappers, and whether it is ignored or mapped is wrapper-dependent. Custom wrappers should accept or ignore unknown flags to avoid breaking when `--model` is present | empty |

| `review_model` | Model for review phases. Falls back to `task_model` if empty. Intended for the default Claude CLI; when set, the executor may still pass `--model` to custom wrappers, and whether it is ignored or mapped is wrapper-dependent. Custom wrappers should accept or ignore unknown flags to avoid breaking when `--model` is present | empty |

Copilot · 2026-04-16T22:51:09Z

+// stripFlag removes a flag and its value from args (e.g., stripFlag(args, "--model")
+// removes both "--model" and the following element). Returns a new slice.
+func stripFlag(args []string, flag string) []string {
+	result := make([]string, 0, len(args))
+	for i := 0; i < len(args); i++ {
+		if args[i] == flag && i+1 < len(args) {
+			i++ // skip flag and its value
+			continue
+		}
+		result = append(result, args[i])


stripFlag only removes the separate "--model " form when a value is present. It does not remove "--model=" (common long-flag form) and it leaves a trailing "--model" with no value intact, which can still produce duplicate/conflicting --model flags or a CLI parse error. Consider stripping both "--model=<...>" and a bare "--model" (no value) as well as repeated occurrences.

Suggested change

// stripFlag removes a flag and its value from args (e.g., stripFlag(args, "--model")

// removes both "--model" and the following element). Returns a new slice.

func stripFlag(args []string, flag string) []string {

result := make([]string, 0, len(args))

for i := 0; i < len(args); i++ {

if args[i] == flag && i+1 < len(args) {

i++ // skip flag and its value

continue

}

result = append(result, args[i])

// stripFlag removes all occurrences of a flag from args.

// It strips both "--flag value" and "--flag=value" forms, and also removes

// a bare trailing "--flag" with no value. Returns a new slice.

func stripFlag(args []string, flag string) []string {

result := make([]string, 0, len(args))

for i := 0; i < len(args); i++ {

switch {

case args[i] == flag:

if i+1 < len(args) {

i++ // skip flag and its value when provided separately

}

continue

case strings.HasPrefix(args[i], flag+"="):

continue // skip --flag=value form

default:

result = append(result, args[i])

}

Copilot · 2026-04-16T22:51:10Z

+func TestStripFlag(t *testing.T) {
+	tests := []struct {
+		name string
+		args []string
+		flag string
+		want []string
+	}{
+		{name: "removes flag and value", args: []string{"--verbose", "--model", "opus", "--print"}, flag: "--model", want: []string{"--verbose", "--print"}},
+		{name: "flag not present", args: []string{"--verbose", "--print"}, flag: "--model", want: []string{"--verbose", "--print"}},
+		{name: "flag at end with value", args: []string{"--verbose", "--model", "opus"}, flag: "--model", want: []string{"--verbose"}},
+		{name: "empty args", args: []string{}, flag: "--model", want: []string{}},
+	}


The new stripFlag tests don’t cover the problematic cases that currently slip through: "--model=opus" and a bare "--model" without a value. Adding these cases would lock in the intended behavior and prevent regressions when users include long flags in claude_args.

Copilot · 2026-04-16T22:51:10Z

 	ReviewPatience        int           `long:"review-patience" default:"0" description:"terminate external review after N unchanged rounds (0 = disabled)"`
-	ClaudeModel           string        `long:"claude-model" description:"model for task execution (e.g., opus, sonnet, haiku)"`
-	ReviewModel           string        `long:"review-model" description:"model for review phases (falls back to --claude-model)"`
+	TaskModel             string        `long:"task-model" description:"model for task execution (e.g., opus, sonnet, haiku)"`


The CLI flag was renamed from --claude-model to --task-model without an alias. If users already adopted --claude-model from the partially-merged feature, their scripts will break. Consider keeping --claude-model as a deprecated alias (with lower precedence than --task-model when both are provided) for at least one release cycle.

Suggested change

TaskModel string `long:"task-model" description:"model for task execution (e.g., opus, sonnet, haiku)"`

TaskModel string `long:"task-model" description:"model for task execution (e.g., opus, sonnet, haiku)"`

ClaudeModel string `long:"claude-model" hidden:"true" description:"deprecated alias for --task-model"`

Copilot · 2026-04-16T22:51:10Z


 **Stalemate detection:** `review_patience` config option (or `--review-patience` CLI flag) terminates the external review loop early when Claude produces no commits for N consecutive rounds. Set to 0 (default) to disable. Useful when the external tool and Claude can't agree on findings.

+**Per-phase model configuration:** `task_model` config option (or `--task-model` CLI flag) sets the model for task execution (e.g., `opus`, `sonnet`, `haiku`). `review_model` (or `--review-model`) sets the model for review phases; falls back to `task_model` if empty. When set, `--model <value>` is injected into the Claude CLI args. Only applies to the default Claude CLI — custom wrappers (`claude_command`) ignore the `--model` flag. Empty by default (uses Claude CLI's default model).


This doc states model overrides are ignored by custom wrappers, but the executor always appends [--model ] whenever task_model/review_model is set, regardless of claude_command. Whether it is ignored or mapped is wrapper-dependent (the wrapper may choose to consume/translate it). Consider rephrasing to reflect that ralphex passes --model and wrappers may ignore or implement it.

Suggested change

**Per-phase model configuration:** `task_model` config option (or `--task-model` CLI flag) sets the model for task execution (e.g., `opus`, `sonnet`, `haiku`). `review_model` (or `--review-model`) sets the model for review phases; falls back to `task_model` if empty. When set, `--model <value>` is injected into the Claude CLI args. Only applies to the default Claude CLI — custom wrappers (`claude_command`) ignore the `--model` flag. Empty by default (uses Claude CLI's default model).

**Per-phase model configuration:** `task_model` config option (or `--task-model` CLI flag) sets the model for task execution (e.g., `opus`, `sonnet`, `haiku`). `review_model` (or `--review-model`) sets the model for review phases; falls back to `task_model` if empty. When set, ralphex appends `--model <value>` to the Claude command args. With the default Claude CLI, this selects the requested model. With custom wrappers (`claude_command`), handling of `--model` is wrapper-dependent: wrappers may ignore it, consume it directly, or translate it to their own equivalent. Empty by default (uses Claude CLI's default model).

- stripFlag now handles --flag=value and bare --flag forms (was only handling --flag value space-form), prevents duplicate/orphaned --model in claude_args - add stripFlag tests for equals form, bare flag, repeated occurrences, prefix-only false-match - doc wording: clarify that task_model/review_model are passed as --model to the configured claude_command and may be ignored or implemented by custom wrappers

Copilot AI review requested due to automatic review settings April 16, 2026 22:46

Copilot started reviewing on behalf of umputun April 16, 2026 22:47 View session

Copilot AI reviewed Apr 16, 2026

View reviewed changes

umputun merged commit b748eb5 into master Apr 16, 2026
2 of 3 checks passed

umputun deleted the task-model-rename branch April 16, 2026 22:56

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Uh oh!

feat: rename claude_model to task_model, address unmerged #272 review findings#284

feat: rename claude_model to task_model, address unmerged #272 review findings#284
umputun merged 2 commits intomasterfrom
task-model-rename

umputun commented Apr 16, 2026

Uh oh!

cloudflare-workers-and-pages Bot commented Apr 16, 2026 •

edited

Loading

Uh oh!

Copilot AI left a comment

Uh oh!

Copilot AI Apr 16, 2026

Uh oh!

Copilot AI Apr 16, 2026

Uh oh!

Copilot AI Apr 16, 2026

Uh oh!

Copilot AI Apr 16, 2026

Uh oh!

Copilot AI Apr 16, 2026

Uh oh!

Copilot AI Apr 16, 2026

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants

		\| `task_model` \| Model for task execution (e.g., `opus`, `sonnet`). Only applies to default Claude CLI, ignored by custom wrappers \| empty \|
		\| `review_model` \| Model for review phases. Falls back to `task_model` if empty. Only applies to default Claude CLI \| empty \|

	TaskModel string `long:"task-model" description:"model for task execution (e.g., opus, sonnet, haiku)"`
	TaskModel string `long:"task-model" description:"model for task execution (e.g., opus, sonnet, haiku)"`
	ClaudeModel string `long:"claude-model" hidden:"true" description:"deprecated alias for --task-model"`


		Stalemate detection: `review_patience` config option (or `--review-patience` CLI flag) terminates the external review loop early when Claude produces no commits for N consecutive rounds. Set to 0 (default) to disable. Useful when the external tool and Claude can't agree on findings.

		Per-phase model configuration: `task_model` config option (or `--task-model` CLI flag) sets the model for task execution (e.g., `opus`, `sonnet`, `haiku`). `review_model` (or `--review-model`) sets the model for review phases; falls back to `task_model` if empty. When set, `--model <value>` is injected into the Claude CLI args. Only applies to the default Claude CLI — custom wrappers (`claude_command`) ignore the `--model` flag. Empty by default (uses Claude CLI's default model).

Uh oh!

Conversation

umputun commented Apr 16, 2026

Uh oh!

cloudflare-workers-and-pages Bot commented Apr 16, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Deploying ralphex with Cloudflare Pages

Uh oh!

Copilot AI left a comment

Choose a reason for hiding this comment

Pull request overview

Reviewed changes

Uh oh!

Copilot AI Apr 16, 2026

Choose a reason for hiding this comment

Uh oh!

Copilot AI Apr 16, 2026

Choose a reason for hiding this comment

Uh oh!

Copilot AI Apr 16, 2026

Choose a reason for hiding this comment

Uh oh!

Copilot AI Apr 16, 2026

Choose a reason for hiding this comment

Uh oh!

Copilot AI Apr 16, 2026

Choose a reason for hiding this comment

Uh oh!

Copilot AI Apr 16, 2026

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants

cloudflare-workers-and-pages Bot commented Apr 16, 2026 •

edited

Loading