v0.1.8

What's Changed

• Bottle v0.1.7 by @github-actions[bot] in #177
• Add per-row and pooled latency metrics to eval command by @alexkroman in #178
• Add --llm-reduce: one LLM prompt over all batch-transcribe results by @alexkroman in #179
• Support batch transcript fetching from piped stdin by @alexkroman in #180
• Add batch streaming mode: assembly stream --from-stdin by @alexkroman in #181
• Add --show-code flag to agent-cascade for script generation by @alexkroman in #182
• Document transcript summarization workflow in README by @alexkroman in #183
• Add --llm and --llm-reduce flags to assembly eval by @alexkroman in #184
• Centralize stdio TTY checks into dedicated functions by @alexkroman in #185
• Add --save-audio flag to record streamed PCM to WAV by @alexkroman in #186
• Add dependency-free diarization error rate (DER) scoring by @alexkroman in #187
• Support multiple datasets in eval command by @alexkroman in #188
• Add file argument support to llm command by @alexkroman in #189
• Simplify update notifier helpers by @alexkroman in #190
• Add --save-transcript and --save-dir flags for transcript/audio file output by @alexkroman in #191
• Add -o/--output flag to dictate command for text/json modes by @alexkroman in #192
• Add podcast RSS/Atom feed expansion to transcribe command by @alexkroman in #195
• Stop streaming commands cleanly on SIGTERM, like Ctrl-C by @alexkroman in #194
• Support per-channel WAV files for --system-audio with --save-dir by @alexkroman in #196
• Support piped stdout in dictate: auto-start single utterance by @alexkroman in #193
• Add guards for stale help examples and JSON output purity by @alexkroman in #197
• Respect --quiet flag for reduce progress table by @alexkroman in #198
• Add field projection (-o) to read commands for pipe-friendly column output by @alexkroman in #199
• Map Ctrl-C to exit code 130 (cancel) instead of 0 by @alexkroman in #202
• Add --url flag to speak command for reading web pages aloud by @alexkroman in #201
• Allow variadic positional sources for hand-picked batch transcription by @alexkroman in #203
• Increase default max-tokens from 1000 to 8192 by @alexkroman in #204
• Add --save-dir finalization: auto-name, note, and sidecar by @alexkroman in #200
• Make login default to production, not stored env by @alexkroman in #205
• Simplify dictate to auto-start recording and exit after one utterance by @alexkroman in #206
• Fix error handling and edge cases across auth, keys, and agent systems by @alexkroman in #207
• llm: accept a directory argument, recursing for .md/.txt by @alexkroman in #209
• fix(telemetry): don't record a cancelled command as an error by @alexkroman in #210
• Simplify: dedup actor label, reuse JSON policy, StrEnum-ify usage window by @alexkroman in #211
• refactor(voices): extract shared --list-voices rendering and voice validation by @alexkroman in #212
• Restrict the sandbox to AssemblyAI logins by @alexkroman in #208
• feat(clip,dub,caption): add --from-stdin batch mode by @alexkroman in #213
• Refactor eval and stream exec modules below the size limit by @alexkroman in #214
• Extract the OS keyring into core/keyring_store by @alexkroman in #215
• refactor(code_gen): type LLM-Gateway options to drop cast() escape hatches by @alexkroman in #216
• Serialize config.toml writes with a cross-process lock (+ concurrency tests) by @alexkroman in #217

Full Changelog: v0.1.7...v0.1.8

v0.1.7

What's Changed

• Bottle v0.1.6 by @github-actions[bot] in #166
• Simplify transcribe/config-builder duplication by @alexkroman in #167
• Simplify dictate language parsing and hoist help command rank by @alexkroman in #168
• Add --turn-detection presets to assembly stream by @alexkroman in #169
• Add cascaded agent-framework init template; make all init templates importable + type-checked by @alexkroman in #170
• Add assembly agent-framework: a terminal client-orchestrated voice cascade by @alexkroman in #171
• Use preserve_logging_state in test_ws websockets-silencer tests by @alexkroman in #172
• Extract shared system-prompt resolver for the two voice commands by @alexkroman in #173
• Document cross-platform portability and release gotchas in AGENTS.md by @alexkroman in #174
• Rename agent-framework command and module to agent-cascade by @alexkroman in #175
• Fix agent-cascade/speak streaming TTS and thread STT/LLM/TTS options by @alexkroman in #176

Full Changelog: v0.1.6...v0.1.7

v0.1.6

What's Changed

• Bottle v0.1.5 by @github-actions[bot] in #123
• Extract WebSocket authorization logic to shared diagnostics module by @alexkroman in #124
• Add opt-in diagnostic logging behind -v/--verbose flag by @alexkroman in #125
• Add VTT subtitle export and --chars-per-caption option by @alexkroman in #126
• Collapse websockets binary frame hex dumps to byte counts by @alexkroman in #127
• Add karaoke subtitle burning example to README by @alexkroman in #128
• Add assembly clip command to cut media by transcript content by @alexkroman in #129
• Refresh uv.lock and adopt click 8.4 in the show-code gate by @alexkroman in #130
• Snap clip boundaries into detected silence to avoid mid-word cuts by @alexkroman in #132
• Split CodeQL Swift analysis into separate workflow by @alexkroman in #133
• Add assembly dictate: hotkey-driven push-to-talk transcription by @alexkroman in #131
• Add fsspec-backed remote storage support for bucket URLs by @alexkroman in #134
• Add assembly dub command for video/audio dubbing by @alexkroman in #135
• Switch speak/dub default voices to the requested language's native voice by @alexkroman in #136
• Split init/dev/share/deploy/eval into options/run exec modules by @alexkroman in #138
• Add video support (--video) to clip/dub and a new caption command by @alexkroman in #139
• Update README with new caption command and improved examples by @alexkroman in #141
• Add --download-sections flag to dub command for partial video dubbing by @alexkroman in #140
• Deduplicate clip/dub media scaffolding into shared mediafile module by @alexkroman in #137
• Refactor README in the style of leading CLI READMEs by @alexkroman in #142
• Deduplicate FakeYoutubeDL test doubles into a shared stub factory by @alexkroman in #144
• Restructure for concurrent agent contributions: derive registration, isolate patches, split guidance by @alexkroman in #145
• Adopt UX/DX conventions from the Codex/Gemini/Antigravity CLIs by @alexkroman in #146
• tests: pin render width suite-wide so --help asserts match CI by @alexkroman in #147
• Fold per-command run-logic into commands// packages by @alexkroman in #148
• Simplify: drop vestigial local re-imports and collapse is_agentic by @alexkroman in #149
• fix: render the [sandbox] tag on speak/dub help instead of eating it by @alexkroman in #150
• Reduce LOC with shared media-source, JSON-emit, and omit-None helpers by @alexkroman in #151
• Tighten the gate: add ASYNC/LOG/TID/etc. ruff rules + remaining mypy strict flags by @alexkroman in #153
• Refactor near-500-line files along natural seams by @alexkroman in #152
• test: drop no-op resolve_json monkeypatching by @alexkroman in #155
• Refactor CLI help copy to terse, period-less Codex-CLI style by @alexkroman in #154
• Restructure aai_cli into a layered package (commands > app > ui > core) by @alexkroman in #156
• Type the agent NDJSON events and HF /splits parsing with Pydantic by @alexkroman in #157
• Narrow handshake_status with isinstance instead of int() coercion by @alexkroman in #158
• Drop redundant self-path header comments by @alexkroman in #159
• Let release.yml be cut from a manual workflow_dispatch by @alexkroman in #160
• Type llm messages with ChatCompletionMessageParam instead of suppressing by @alexkroman in #161
• Fix CLI error-handling papercuts and make the gate pass on macOS by @alexkroman in #163
• ci: add Windows to the test matrix and fix Windows-only test breakage by @alexkroman in #162
• Fix CLI QA papercuts: silent --download-sections drop, port bounds, init exit codes by @alexkroman in #164
• Fix release.yml tag job: set git identity for annotated tag by @alexkroman in #165

Full Changelog: v0.1.5...v0.1.6

v0.1.5

What's Changed

• gitleaks: tolerate newer versions (allowlist a generic-api-key false positive) by @alexkroman in #114
• Bottle v0.1.4 by @github-actions[bot] in #115
• Split CLI help snapshots by command group to reduce merge conflicts by @alexkroman in #116
• Derive version from git tags via hatch-vcs by @alexkroman in #117

Full Changelog: v0.1.4...v0.1.5

v0.1.4

What's Changed

• Consolidate error classification for realtime WebSocket failures by @alexkroman in #107
• Bottle v0.1.3 by @github-actions[bot] in #108
• Refactor usage models to Pydantic with graceful degradation by @alexkroman in #109
• Add assembly webhooks listen command for local webhook development by @alexkroman in #110
• Refactor context module: extract utilities, consolidate error handling by @alexkroman in #112
• Refresh README with updated feature descriptions and examples by @alexkroman in #111
• Bump version to 0.1.4 by @alexkroman in #113

Full Changelog: v0.1.3...v0.1.4

v0.1.3

What's Changed

• Add Swift CodeQL analysis for macOS system audio helper by @alexkroman in #103
• Fix all 38 CodeQL quality alerts and gate CodeQL locally in check.sh by @alexkroman in #104
• Drop DER scoring and the pyannote-metrics dependency by @alexkroman in #105
• Bump version to 0.1.3 by @alexkroman in #106

Full Changelog: v0.1.2...v0.1.3

v0.1.0

What's Changed

• Add AssemblyAI CLI (aai) by @alexkroman in #1
• Build(deps): Bump the actions group with 2 updates by @dependabot[bot] in #2
• feat(cli): unify transcribe + stream LLM Gateway under --llm by @alexkroman in #5
• Build(deps): Bump the python-deps group with 16 updates by @dependabot[bot] in #4
• feat(auth): browser-based Stytch OAuth login + environment switching by @alexkroman in #6
• feat: aai init — interactive project scaffolder (transcribe, stream, agent) by @alexkroman in #7
• Help examples, error suggestions, and test cleanup by @alexkroman in #8
• Install-path testing: install.sh smoke tests, CI, and dev docs by @alexkroman in #9
• Remove dead code with no production callers by @alexkroman in #10
• Improve Claude Code setup: SessionStart hook, git perms, planning + review wiring by @alexkroman in #12
• feat(config): persist Stytch session + account id per profile by @alexkroman in #11
• simplify: dedupe webhook-header, code-gen gateway, proc-detail, and nested-attr helpers by @alexkroman in #13
• refactor: centralize resolution, unify auth detection, harden agent/config by @alexkroman in #14
• Narrow bare dict/list annotations to parameterized types by @alexkroman in #15
• Env-aware Voice Agent URL + grouped --help and Vercel/Supabase-style copy, colors & UX by @alexkroman in #16
• chore: strengthen CLI quality gates by @alexkroman in #17
• Add macOS system audio streaming by @alexkroman in #18
• Leverage pydantic, simplify shared helpers, and lint CI workflows by @alexkroman in #19
• Refactor README into Supabase/Vercel CLI house style by @alexkroman in #20
• Add bundled aai-cli skill and aai setup; adopt AGENTS.md by @alexkroman in #21
• Unify CLI tables on one minimal, consistent style by @alexkroman in #22
• test: cover human-render, init launch, and error-path gaps by @alexkroman in #23
• Clarify transcribe/stream CLI help text by @alexkroman in #24
• Homebrew tap, Python 3.12 floor, and install/CI hardening by @alexkroman in #26
• docs/config: sync gate docs to check.sh; expand agent allowlist by @alexkroman in #28
• Remove code duplication across CLI modules by @alexkroman in #27
• Tighten complexity/length gates and refactor to satisfy them by @alexkroman in #29
• Add actionlint, zizmor, gitleaks, and refurb gates to check.sh by @alexkroman in #30
• test: kill surviving mutants in errors, timeparse, follow by @alexkroman in #31
• Add state nonce to OAuth flow for login CSRF protection by @alexkroman in #32
• Harden YouTube download fallback to pick the largest landed file by @alexkroman in #34
• Route root-callback env errors through the standard error path by @alexkroman in #35
• Improve CLI UX: enums, validation, help panels, and completions by @alexkroman in #33
• Tidy a few non-idiomatic patterns by @alexkroman in #36
• Split oversized modules and gate file length at 500 lines by @alexkroman in #38
• Make CLI output idioms match the tools agents see most by @alexkroman in #37
• Add Typer range validation to numeric CLI options by @alexkroman in #39
• Refactor: extract shared muted-footnote/stack render helpers by @alexkroman in #40
• Clear mypy annotation-unchecked notes in tests by @alexkroman in #41
• Fix license badge and simplify README prose by @alexkroman in #43
• Test brew/pipx/uv tool installs from the branch in CI by @alexkroman in #42
• Default to production environment; drop login state nonce by @alexkroman in #44
• Add stream --llm-interval throttle and fix show-code LLM-chain parity by @alexkroman in #45
• Add aai onboard guided onboarding wizard by @alexkroman in #46
• test: standardize SDK-boundary mocking on pytest-mock by @alexkroman in #47
• Simplify: dedupe optional-key resolution, status cells, doctor checks by @alexkroman in #48
• [Aikido] AI Fix for Potential file inclusion attack via reading file by @aikido-autofix[bot] in #49
• Add aai dev/share/deploy and make starter templates deployable beyond Vercel by @alexkroman in #51
• Restyle aai init starter templates with the AssemblyAI brand system by @alexkroman in #52
• Lazy-validate AAI_AUTH_PORT to prevent import-time crashes by @alexkroman in #53
• Robustness, environment-aware code gen, and config caching by @alexkroman in #54
• Validate input before credentials; fix help rendering and closed pipes by @alexkroman in #55
• Fix QA findings: non-interactive auth, input validation, error hygiene by @alexkroman in #56
• test: kill surviving mutants in config/output/microphone/transcribe_render by @alexkroman in #57
• Fix UX/DX audit findings across the CLI surface by @alexkroman in #58
• Fix YouTube --show-code, Vercel deploy framework, stream poll default, --llm help by @alexkroman in #59
• Add record/replay fixture harness for end-to-end CLI tests by @alexkroman in #60
• Add sandbox-only aai speak streaming TTS command by @alexkroman in #61
• Fix code-review findings: JSON list output, agent/TTS robustness, auth-flow structure by @alexkroman in #62
• Transcribe and stream podcast pages (any yt-dlp-extractable URL) by @alexkroman in #63
• Add aai transcribe --download-sections (yt-dlp passthrough) by @alexkroman in #64
• Rename CLI command from aai to assembly by @alexkroman in #65
• Add anonymous usage telemetry collection by @alexkroman in #66
• Add Datadog CLI-usage dashboard + JSON validity gate by @alexkroman in #67
• Reorganize code_gen and command tests by concern by @alexkroman in #68
• Ship a prebuilt Homebrew bottle on release; drop install.sh by @alexkroman in #69

New Contributors

• @alexkroman made their first contribution in #1
• @dependabot[bot] made their first contribution in #2
• @aikido-autofix[bot] made their first contribution in #49

Full Changelog: https://github.com/AssemblyAI/cli/commits/v0.1.0