Explicit --format is ignored when --output has a recognized extension

[100yenadmin]

Hi, I hit a format precedence bug in sag 0.4.0.

Repro

sag speak \
  --voice-id "$ELEVENLABS_VOICE_ID" \
  --output out.mp3 \
  --format mp3_44100_192 \
  --no-stream \
  "format precedence check"

ffprobe -v quiet -show_entries format=bit_rate -of default=noprint_wrappers=1:nokey=1 out.mp3

Expected: out.mp3 is requested/generated as mp3_44100_192.

Actual: the saved file is still 128 kbps.

I also verified the same ElevenLabs format directly against the API (output_format=mp3_44100_192) and that returns a true 192 kbps MP3, so the mismatch appears to be in sag's CLI/request construction rather than ElevenLabs.

Root cause

cmd/speak.go currently re-infers the format from --output unconditionally:

• --output out.mp3 resets opts.outputFmt back to the inferred .mp3 default (mp3_44100_128)
• that happens even when --format was explicitly set

That contradicts the docs, which say explicit --format should win.

Proposed fix

Only infer from the output extension when --format was not explicitly provided. I have a small patch + regression test ready and will open a PR right after filing this.

[clawsweeper]

Codex review: keeping this open for maintainer follow-up; there is still a little grit to resolve. Reviewed June 19, 2026, 1:32 PM ET / 17:32 UTC.

Summary
This should stay open: current main still lets output-extension inference overwrite an explicit --format, while the linked pull request is open and should close this issue only if it lands.

Reproducibility: yes. by source inspection: current main overwrites opts.outputFmt from a recognized output extension before request construction, matching the issue's --output out.mp3 --format mp3_44100_192 scenario. I did not run the live ElevenLabs command in this read-only review.

Root-cause cluster
Relationship: canonical
Canonical: #21
Summary: This issue is the canonical report for the explicit-format precedence bug, and the open linked pull request is the candidate fix.

Members:

• fixed_by_candidate: Honor explicit --format over output extension #22 - The pull request uses closing syntax for this issue and changes the same cmd/speak.go extension-inference path.

Proposal only: this assessment does not dispatch repair, suppress jobs, mutate sibling items, close, or merge anything.

Next step
An open linked pull request already owns the fix, so ClawSweeper should not queue a competing repair branch.

Review details

Best possible solution:

Keep this issue as the canonical bug until #22 lands or maintainers choose a replacement fix with the same precedence guarantee and regression coverage.

Do we have a high-confidence way to reproduce the issue?

Yes by source inspection: current main overwrites opts.outputFmt from a recognized output extension before request construction, matching the issue's --output out.mp3 --format mp3_44100_192 scenario. I did not run the live ElevenLabs command in this read-only review.

Is this the best way to solve the issue?

Yes. Guarding extension inference with cmd.Flags().Changed("format"), as the linked PR proposes, is the narrow maintainable fix that matches the docs; the issue should remain open until that fix merges.

AGENTS.md: not found in the target repository.

Remaining risk / open question:

• No live ElevenLabs command was run during this read-only review, so the bitrate observation remains reporter-provided; the source path is still clear enough to keep the issue open.
• The linked pull request is open and has not yet supplied real behavior proof, so this issue should not close ahead of a merged fix.

Codex review notes: model internal, reasoning high; reviewed against ad9b1d3013b4.

Label changes

Label changes:

• remove clawsweeper:needs-maintainer-review: Current issue advisory state no longer selects this label.

Label justifications:

• P2: Explicit audio format selection is ignored in common save-to-file usage, which is a normal CLI behavior bug with limited blast radius.
• impact:other: The issue affects generated audio format and bitrate selection, which is maintainer-visible but outside the specific owned impact classes.

Evidence reviewed

What I checked:

• Current main overwrites explicit format: sag speak calls inferFormatFromExt(opts.outputPath) whenever an output path is present and assigns the result to opts.outputFmt without checking whether the format flag was explicitly changed. (cmd/speak.go:111, ad9b1d3013b4)
• Request uses the overwritten value: The ElevenLabs request is built from opts.outputFmt, so the extension-inferred value is the value sent as OutputFormat. (cmd/speak.go:399, ad9b1d3013b4)
• Docs define explicit format as higher precedence: The format docs describe extension inference and then state that explicit --format <string> always wins. (docs/formats.md:8, ad9b1d3013b4)
• Architecture docs agree on precedence: The architecture docs say inferFormatFromExt maps known extensions and that --format always wins. (docs/architecture.md:99, ad9b1d3013b4)
• Linked fix is still open: Honor explicit --format over output extension #22 is open, references this issue as a closing issue, and proposes guarding extension inference behind !cmd.Flags().Changed("format"). (ead6eff7464f)
• No merged canonical replacement found: Search found this issue and the open linked pull request as the active owner for the format/output-extension problem; related merged PRs are adjacent output-format work, not this precedence fix.

Likely related people:

• steipete: Blame and git log -S tie the current speak command, extension inference, request construction, and format docs to the v0.4.0 implementation history. (role: introduced behavior and recent area contributor; confidence: high; commits: c451f4673401, ad9b1d3013b4; files: cmd/speak.go, docs/formats.md, docs/architecture.md)
• derspotter: A prior merged commit changed output format handling for Opus/OGG in the same command area, making this person relevant context for format-routing behavior. (role: adjacent output-format contributor; confidence: medium; commits: 8e44324fb455; files: cmd/speak.go, cmd/speak_test.go, CHANGELOG.md)

How this review workflow works

• ClawSweeper keeps one durable marker-backed review comment per issue or PR.
• Re-runs edit this comment so the latest verdict, findings, and automation markers stay together instead of adding duplicate bot comments.
• A fresh review can be triggered by eligible @clawsweeper re-review comments, exact-item GitHub events, scheduled/background review runs, or manual workflow dispatch.
• PR/issue authors and users with repository write access can comment @clawsweeper re-review or @clawsweeper re-run on an open PR or issue to request a fresh review only.
• Maintainers can also comment @clawsweeper review to request a fresh review only.
• Fresh-review commands do not start repair, autofix, rebase, CI repair, or automerge.
• Maintainer-only repair and merge flows require explicit commands such as @clawsweeper autofix, @clawsweeper automerge, @clawsweeper fix ci, or @clawsweeper address review.
• Maintainers can comment @clawsweeper explain to ask for more context, or @clawsweeper stop to stop active automation.

[100yenadmin]

cc @steipete - opened PR #22 with the small precedence fix and regression test.