evo-cli 0.21.2__tar.gz → 0.22.0__tar.gz

{evo_cli-0.21.2 → evo_cli-0.22.0}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: evo_cli
-Version: 0.21.2
+Version: 0.22.0
 Summary: Evolution CLI - a developer toolbox for setting up dev machines
 Author: maycuatroi
 Project-URL: Homepage, https://github.com/maycuatroi/evo-cli
@@ -195,53 +195,66 @@ uses `git pull --ff-only` so the command never creates merge commits.
 
 #### Text to Speech
 
-Synthesise speech through Vbee (Vietnamese) or OpenAI `gpt-4o-mini-tts`, and play it right away:
+Synthesise speech through Gemini `gemini-3.1-flash-tts-preview`, Vbee (Vietnamese), or OpenAI
+`gpt-4o-mini-tts`, and play it right away:
 
 ```bash
 evo tts speak "Xin chào, bản build đã xong"
-evo tts speak -f notes.md -o notes.mp3
+evo tts speak -f notes.md -o notes.wav
+evo tts speak "hôm nay trời đẹp" -V Sulafat --instructions "kể chuyện, ấm áp"
+evo tts speak "[whispers] đừng nói với ai nhé"
 evo tts speak "hello there" -p openai -V nova --instructions "calm and encouraging"
 git log -1 --format=%s | evo tts speak
 ```
 
-`speak` is the realtime path. Text longer than the provider's per-request limit (Vbee 300
-characters, OpenAI 4000) is split on sentence boundaries and the audio is joined back into one file,
-so the first words start playing while the rest is still being synthesised.
+`speak` is the realtime path. Text longer than the provider's per-request limit (Gemini 2000
+characters, Vbee 300, OpenAI 4000) is split on sentence boundaries and the audio is joined back into
+one file, so the first words start playing while the rest is still being synthesised.
+
+Gemini is the most expressive of the three: 30 prebuilt voices, automatic language detection across
+70+ languages, free-form delivery notes through `--instructions`, and inline audio tags such as
+`[whispers]`, `[excited]`, `[sighs]`, or `[very slow]` anywhere in the text. It answers with raw
+24 kHz PCM, so `--format wav` is the default there; `--format mp3` re-encodes through `ffmpeg` and
+needs it on PATH. The format is also inferred from the `--output` suffix, so `-o notes.mp3` still
+produces an mp3.
 
 For bulk work use the batch path, which goes through Vbee's async API and polls
 `/v1/tts/requests/{id}` until each audio link appears:
 
 ```bash
-evo tts batch chapters/ -o audio/          # one mp3 per .txt/.md file
+evo tts batch chapters/ -o audio/          # one file per .txt/.md input
 evo tts batch a.txt b.txt -c 8             # 8 items in flight
 evo tts batch --manifest jobs.jsonl        # {"id":.., "text":.., "voice":..} per line
 ```
 
-OpenAI has no batch speech endpoint, so with `-p openai` the items are parallelised locally instead.
+Gemini and OpenAI have no batch speech endpoint, so there the items are parallelised locally instead.
 
-Voice codes come from `evo tts voices` (`-l en-US`, `--gender male`, `-p openai`, `--json`).
+Voice codes come from `evo tts voices` (`-p gemini`, `-p openai`, `-l en-US`, `--gender male`,
+`--json`).
 
 Credentials live in the omelet store, never in flags or source:
 
 ```bash
+evo cred add gemini_api_key --from-stdin   # key from https://aistudio.google.com/apikey
 evo cred add vbee.app_id --from-stdin      # UUID from https://studio.vbee.vn/apps
 evo cred add vbee.token --from-stdin       # JWT from the same app page
 evo cred add openai_api_key --from-stdin
 ```
 
-`VBEE_APP_ID`, `VBEE_TOKEN`, and `OPENAI_API_KEY` override the store when set.
+`GEMINI_API_KEY` (or `GOOGLE_API_KEY`), `VBEE_APP_ID`, `VBEE_TOKEN`, and `OPENAI_API_KEY` override
+the store when set.
 
 `--provider auto` (the default) resolves to `EVO_TTS_PROVIDER` when that is set, and otherwise to
-whichever provider has credentials. Pick a machine default once:
+the first of Gemini, Vbee, OpenAI that has credentials. Pick a machine default once:
 
 ```bash
-export EVO_TTS_PROVIDER=openai        # what `auto` means here
-export EVO_TTS_VOICE_OPENAI=nova      # default voice for that provider only
+export EVO_TTS_PROVIDER=gemini        # what `auto` means here
+export EVO_TTS_VOICE_GEMINI=Sulafat   # default voice for that provider only
 ```
 
-Prefer the provider-scoped `EVO_TTS_VOICE_OPENAI` / `EVO_TTS_VOICE_VBEE` over a bare
-`EVO_TTS_VOICE`: a shared value breaks as soon as you pass `--provider vbee`, because an OpenAI
-voice name is not a Vbee voice code.
+Prefer the provider-scoped `EVO_TTS_VOICE_GEMINI` / `EVO_TTS_VOICE_OPENAI` / `EVO_TTS_VOICE_VBEE`
+over a bare `EVO_TTS_VOICE`: a shared value breaks as soon as you pass `--provider vbee`, because a
+Gemini voice name is not a Vbee voice code.
 
 Playback uses whichever of `ffplay`, `mpv`, `cvlc`, `afplay`, or `paplay`/`aplay` is on PATH, and
 falls back to PowerShell's `MediaPlayer` on Windows. Without any of them the audio is still written

{evo_cli-0.21.2 → evo_cli-0.22.0}/README.md

@@ -167,53 +167,66 @@ uses `git pull --ff-only` so the command never creates merge commits.
 
 #### Text to Speech
 
-Synthesise speech through Vbee (Vietnamese) or OpenAI `gpt-4o-mini-tts`, and play it right away:
+Synthesise speech through Gemini `gemini-3.1-flash-tts-preview`, Vbee (Vietnamese), or OpenAI
+`gpt-4o-mini-tts`, and play it right away:
 
 ```bash
 evo tts speak "Xin chào, bản build đã xong"
-evo tts speak -f notes.md -o notes.mp3
+evo tts speak -f notes.md -o notes.wav
+evo tts speak "hôm nay trời đẹp" -V Sulafat --instructions "kể chuyện, ấm áp"
+evo tts speak "[whispers] đừng nói với ai nhé"
 evo tts speak "hello there" -p openai -V nova --instructions "calm and encouraging"
 git log -1 --format=%s | evo tts speak
 ```
 
-`speak` is the realtime path. Text longer than the provider's per-request limit (Vbee 300
-characters, OpenAI 4000) is split on sentence boundaries and the audio is joined back into one file,
-so the first words start playing while the rest is still being synthesised.
+`speak` is the realtime path. Text longer than the provider's per-request limit (Gemini 2000
+characters, Vbee 300, OpenAI 4000) is split on sentence boundaries and the audio is joined back into
+one file, so the first words start playing while the rest is still being synthesised.
+
+Gemini is the most expressive of the three: 30 prebuilt voices, automatic language detection across
+70+ languages, free-form delivery notes through `--instructions`, and inline audio tags such as
+`[whispers]`, `[excited]`, `[sighs]`, or `[very slow]` anywhere in the text. It answers with raw
+24 kHz PCM, so `--format wav` is the default there; `--format mp3` re-encodes through `ffmpeg` and
+needs it on PATH. The format is also inferred from the `--output` suffix, so `-o notes.mp3` still
+produces an mp3.
 
 For bulk work use the batch path, which goes through Vbee's async API and polls
 `/v1/tts/requests/{id}` until each audio link appears:
 
 ```bash
-evo tts batch chapters/ -o audio/          # one mp3 per .txt/.md file
+evo tts batch chapters/ -o audio/          # one file per .txt/.md input
 evo tts batch a.txt b.txt -c 8             # 8 items in flight
 evo tts batch --manifest jobs.jsonl        # {"id":.., "text":.., "voice":..} per line
 ```
 
-OpenAI has no batch speech endpoint, so with `-p openai` the items are parallelised locally instead.
+Gemini and OpenAI have no batch speech endpoint, so there the items are parallelised locally instead.
 
-Voice codes come from `evo tts voices` (`-l en-US`, `--gender male`, `-p openai`, `--json`).
+Voice codes come from `evo tts voices` (`-p gemini`, `-p openai`, `-l en-US`, `--gender male`,
+`--json`).
 
 Credentials live in the omelet store, never in flags or source:
 
 ```bash
+evo cred add gemini_api_key --from-stdin   # key from https://aistudio.google.com/apikey
 evo cred add vbee.app_id --from-stdin      # UUID from https://studio.vbee.vn/apps
 evo cred add vbee.token --from-stdin       # JWT from the same app page
 evo cred add openai_api_key --from-stdin
 ```
 
-`VBEE_APP_ID`, `VBEE_TOKEN`, and `OPENAI_API_KEY` override the store when set.
+`GEMINI_API_KEY` (or `GOOGLE_API_KEY`), `VBEE_APP_ID`, `VBEE_TOKEN`, and `OPENAI_API_KEY` override
+the store when set.
 
 `--provider auto` (the default) resolves to `EVO_TTS_PROVIDER` when that is set, and otherwise to
-whichever provider has credentials. Pick a machine default once:
+the first of Gemini, Vbee, OpenAI that has credentials. Pick a machine default once:
 
 ```bash
-export EVO_TTS_PROVIDER=openai        # what `auto` means here
-export EVO_TTS_VOICE_OPENAI=nova      # default voice for that provider only
+export EVO_TTS_PROVIDER=gemini        # what `auto` means here
+export EVO_TTS_VOICE_GEMINI=Sulafat   # default voice for that provider only
 ```
 
-Prefer the provider-scoped `EVO_TTS_VOICE_OPENAI` / `EVO_TTS_VOICE_VBEE` over a bare
-`EVO_TTS_VOICE`: a shared value breaks as soon as you pass `--provider vbee`, because an OpenAI
-voice name is not a Vbee voice code.
+Prefer the provider-scoped `EVO_TTS_VOICE_GEMINI` / `EVO_TTS_VOICE_OPENAI` / `EVO_TTS_VOICE_VBEE`
+over a bare `EVO_TTS_VOICE`: a shared value breaks as soon as you pass `--provider vbee`, because a
+Gemini voice name is not a Vbee voice code.
 
 Playback uses whichever of `ffplay`, `mpv`, `cvlc`, `afplay`, or `paplay`/`aplay` is on PATH, and
 falls back to PowerShell's `MediaPlayer` on Windows. Without any of them the audio is still written

evo_cli-0.22.0/evo_cli/VERSION

@@ -0,0 +1 @@
+0.22.0

{evo_cli-0.21.2 → evo_cli-0.22.0}/evo_cli/commands/opencode.py

@@ -447,10 +447,13 @@ def configure_opencode_project(project_path):
 def verify_mcp_servers():
     """Run a basic JSON-RPC initialize check against installed MCP servers."""
     step("Verifying MCP servers")
+    # MCP frames stdio messages one per line, so without the trailing newline the
+    # server sees an unterminated frame, exits 0 on EOF and answers nothing -
+    # which read as "server broken" for every server we ever verified.
     init_message = (
         '{"jsonrpc":"2.0","id":1,"method":"initialize",'
         '"params":{"protocolVersion":"2024-11-05","capabilities":{},'
-        '"clientInfo":{"name":"evo-cli","version":"1.0"}}}'
+        '"clientInfo":{"name":"evo-cli","version":"1.0"}}}\n'
     )
     for name, cmd in _local_mcp_commands():
         try:

{evo_cli-0.21.2 → evo_cli-0.22.0}/evo_cli/commands/tts.py

@@ -16,10 +16,12 @@ TEXT_SUFFIXES = (".txt", ".md")
 SPEAK_EPILOG = Text.from_markup(
     "[bold]Examples[/bold]\n\n"
     "  [cyan]evo tts speak 'Xin chào, bản build đã xong'[/cyan]   speak it out loud now\n"
-    "  [cyan]evo tts speak -f notes.md -o notes.mp3[/cyan]        read a file, keep the audio\n"
+    "  [cyan]evo tts speak -f notes.md -o notes.wav[/cyan]        read a file, keep the audio\n"
+    "  [cyan]evo tts speak 'hi' -V Sulafat --instructions 'kể chuyện, ấm áp'[/cyan]  steer the delivery\n"
+    "  [cyan]evo tts speak '[whispers] bí mật nhé'[/cyan]         Gemini audio tags work inline\n"
     "  [cyan]evo tts speak 'hello' -p openai -V nova[/cyan]       use gpt-4o-mini-tts instead\n"
     "  [cyan]git log -1 --format=%s | evo tts speak[/cyan]        read stdin\n"
-    "  [cyan]evo tts speak 'hi' --stdout > out.mp3[/cyan]         pipe raw audio"
+    "  [cyan]evo tts speak 'hi' --stdout > out.wav[/cyan]         pipe raw audio"
 )
 
 BATCH_EPILOG = Text.from_markup(
@@ -33,8 +35,9 @@ BATCH_EPILOG = Text.from_markup(
 
 VOICES_EPILOG = Text.from_markup(
     "[bold]Examples[/bold]\n\n"
-    "  [cyan]evo tts voices[/cyan]                                Vbee Vietnamese voices\n"
-    "  [cyan]evo tts voices -l en-US --gender male[/cyan]         filter by language and gender\n"
+    "  [cyan]evo tts voices[/cyan]                                voices of the auto-picked provider\n"
+    "  [cyan]evo tts voices -p gemini[/cyan]                      the 30 Gemini prebuilt voices\n"
+    "  [cyan]evo tts voices -p vbee -l en-US --gender male[/cyan] filter by language and gender\n"
     "  [cyan]evo tts voices -p openai[/cyan]                      gpt-4o-mini-tts voices\n"
     "  [cyan]evo tts voices --json[/cyan]                         machine-readable"
 )
@@ -90,6 +93,16 @@ def collect_items(inputs, texts, manifest):
     return [item for item in items if item["text"].strip()]
 
 
+def resolve_output_format(provider, output_format, output=None):
+    if output_format:
+        return output_format
+    if output:
+        suffix = Path(output).suffix.lstrip(".").lower()
+        if suffix in core.supported_formats(provider):
+            return suffix
+    return core.default_format(provider)
+
+
 def unique_path(out_dir, name, output_format):
     candidate = out_dir / f"{name}.{output_format}"
     counter = 2
@@ -101,12 +114,16 @@ def unique_path(out_dir, name, output_format):
 
 @click.group("tts")
 def tts_group():
-    """**Text to speech** via Vbee (Vietnamese) or OpenAI `gpt-4o-mini-tts`.
+    """**Text to speech** via Gemini `gemini-3.1-flash-tts-preview`, Vbee, or OpenAI `gpt-4o-mini-tts`.
 
     `speak` is the realtime path: it synthesises and plays immediately.
     `batch` is the bulk path: it uses Vbee's async API and writes one file per input.
-    Credentials come from the omelet store (`evo cred add vbee.app_id`, `vbee.token`,
-    `openai_api_key`); nothing is read from hardcoded values.
+    Credentials come from the omelet store (`evo cred add gemini_api_key`, `vbee.app_id`,
+    `vbee.token`, `openai_api_key`); nothing is read from hardcoded values.
+
+    Gemini is the default when its key is stored: 30 expressive voices, any of the
+    70+ supported languages, plus inline audio tags like `[whispers]` or `[excited]`
+    and free-form delivery notes through `--instructions`.
     """
 
 
@@ -116,25 +133,24 @@ def tts_group():
 @click.option(
     "-p",
     "--provider",
-    type=click.Choice(["auto", "vbee", "openai"]),
+    type=click.Choice(["auto", "gemini", "vbee", "openai"]),
     default="auto",
     show_default=True,
-    help="auto picks Vbee when its credentials exist, else OpenAI.",
+    help="auto picks Gemini when its key exists, else Vbee, else OpenAI.",
 )
 @click.option("-V", "--voice", help="Voice code (see `evo tts voices`).")
 @click.option("-o", "--output", help="Keep the audio at this path instead of a temp file.")
 @click.option(
     "--format",
     "output_format",
-    type=click.Choice(["mp3", "wav"]),
-    default="mp3",
-    show_default=True,
-    help="Audio container.",
+    type=click.Choice(["mp3", "wav", "pcm"]),
+    default=None,
+    help="Audio container. Default: wav on Gemini, mp3 elsewhere, or the suffix of --output.",
 )
 @click.option("--speed", type=float, default=1.0, show_default=True, help="Speaking rate (Vbee: 0.25-1.9).")
-@click.option("--bitrate", type=int, default=128, show_default=True, help="Vbee bitrate in kbps.")
-@click.option("--instructions", help="OpenAI only: how the voice should deliver the text.")
-@click.option("--model", help="OpenAI model override (default gpt-4o-mini-tts).")
+@click.option("--bitrate", type=int, default=128, show_default=True, help="Bitrate in kbps (Vbee, Gemini mp3).")
+@click.option("--instructions", help="Gemini/OpenAI: how the voice should deliver the text.")
+@click.option("--model", help="Model override (default gemini-3.1-flash-tts-preview / gpt-4o-mini-tts).")
 @click.option("--no-play", is_flag=True, help="Synthesise only; do not play through the speakers.")
 @click.option("--stdout", "to_stdout", is_flag=True, help="Write raw audio bytes to stdout (implies --no-play).")
 @click.option("-q", "--quiet", is_flag=True, help="Suppress progress output.")
@@ -156,14 +172,15 @@ def speak(
     """Synthesise **TEXT** and play it right away.
 
     Long text is split on sentence boundaries so each request stays inside the
-    provider's realtime limit (Vbee 300 characters, OpenAI 4000), then the
-    resulting audio is joined back into one file.
+    provider's realtime limit (Gemini 2000 characters, Vbee 300, OpenAI 4000),
+    then the resulting audio is joined back into one file.
     """
     body = read_input_text(text, text_file)
     if not quiet and not to_stdout:
         step("evo tts speak")
     try:
         resolved = core.resolve_provider(provider)
+        output_format = resolve_output_format(resolved, output_format, output)
         chunks = core.chunk_limit(resolved, "realtime")
         if not quiet and not to_stdout:
             info(
@@ -221,10 +238,10 @@ def speak(
 @click.option(
     "-p",
     "--provider",
-    type=click.Choice(["auto", "vbee", "openai"]),
+    type=click.Choice(["auto", "gemini", "vbee", "openai"]),
     default="auto",
     show_default=True,
-    help="auto picks Vbee when its credentials exist, else OpenAI.",
+    help="auto picks Gemini when its key exists, else Vbee, else OpenAI.",
 )
 @click.option(
     "--mode",
@@ -237,15 +254,14 @@ def speak(
 @click.option(
     "--format",
     "output_format",
-    type=click.Choice(["mp3", "wav"]),
-    default="mp3",
-    show_default=True,
-    help="Audio container.",
+    type=click.Choice(["mp3", "wav", "pcm"]),
+    default=None,
+    help="Audio container. Default: wav on Gemini, mp3 elsewhere.",
 )
 @click.option("--speed", type=float, default=1.0, show_default=True, help="Speaking rate.")
-@click.option("--bitrate", type=int, default=128, show_default=True, help="Vbee bitrate in kbps.")
-@click.option("--instructions", help="OpenAI only: how the voice should deliver the text.")
-@click.option("--model", help="OpenAI model override (default gpt-4o-mini-tts).")
+@click.option("--bitrate", type=int, default=128, show_default=True, help="Bitrate in kbps (Vbee, Gemini mp3).")
+@click.option("--instructions", help="Gemini/OpenAI: how the voice should deliver the text.")
+@click.option("--model", help="Model override (default gemini-3.1-flash-tts-preview / gpt-4o-mini-tts).")
 @click.option("-c", "--concurrency", type=int, default=4, show_default=True, help="Items in flight at once.")
 @click.option("--webhook", help="Vbee webhookUrl; the API requires one even though evo polls for the result.")
 @click.option("--timeout", type=int, default=900, show_default=True, help="Seconds to wait per async request.")
@@ -287,9 +303,10 @@ def batch(
         error(str(exc))
         sys.exit(1)
 
+    output_format = resolve_output_format(resolved, output_format)
     effective_mode = mode
-    if resolved == "openai" and mode == "batch":
-        info("OpenAI has no batch speech endpoint - running the items concurrently instead.")
+    if resolved in ("openai", "gemini") and mode == "batch":
+        info(f"{resolved} has no batch speech endpoint - running the items concurrently instead.")
         effective_mode = "realtime"
 
     target_dir = Path(out_dir)
@@ -340,7 +357,7 @@ def batch(
 @click.option(
     "-p",
     "--provider",
-    type=click.Choice(["auto", "vbee", "openai"]),
+    type=click.Choice(["auto", "gemini", "vbee", "openai"]),
     default="auto",
     show_default=True,
     help="Which catalog to list.",

{evo_cli-0.21.2 → evo_cli-0.22.0}/evo_cli/credentials/registry.py

@@ -71,6 +71,17 @@ SPECS = [
         "rotate": "https://studio.vbee.vn/apps -> open the app -> copy App ID + Token",
         "keys": ["vbee"],
     },
+    {
+        "path": "ai/gemini.json",
+        "id": "gemini",
+        "service": "Google Gemini API",
+        "category": "ai",
+        "type": "api_key",
+        "lifetime": "stable",
+        "description": "Gemini API key (evo tts gemini provider, generateContent)",
+        "rotate": "https://aistudio.google.com/apikey -> create or rotate the key",
+        "keys": ["gemini_api_key"],
+    },
     {
         "path": "ai/google.json",
         "id": "google_api",

{evo_cli-0.21.2 → evo_cli-0.22.0}/evo_cli/tts/__init__.py

@@ -2,6 +2,7 @@ from evo_cli.tts.core import (
     MODES,
     PROVIDERS,
     chunk_limit,
+    default_format,
     default_voice,
     default_voice_for,
     list_voices,
@@ -18,6 +19,7 @@ __all__ = [
     "PROVIDERS",
     "TtsError",
     "chunk_limit",
+    "default_format",
     "default_voice",
     "default_voice_for",
     "list_voices",

{evo_cli-0.21.2 → evo_cli-0.22.0}/evo_cli/tts/core.py

@@ -1,13 +1,14 @@
 import os
 from concurrent.futures import ThreadPoolExecutor, as_completed
 
+from evo_cli.tts import gemini as gemini_tts
 from evo_cli.tts import openai as openai_tts
 from evo_cli.tts import vbee
 from evo_cli.tts.chunking import join_audio, split_text
-from evo_cli.tts.creds import has_openai_credentials, has_vbee_credentials
+from evo_cli.tts.creds import has_gemini_credentials, has_openai_credentials, has_vbee_credentials
 from evo_cli.tts.errors import TtsError
 
-PROVIDERS = ("vbee", "openai")
+PROVIDERS = ("gemini", "vbee", "openai")
 JOINABLE_FORMATS = ("mp3", "wav", "pcm")
 MODES = ("realtime", "batch")
 
@@ -21,12 +22,15 @@ def resolve_provider(provider):
             return preferred
         if preferred and preferred != "auto":
             raise TtsError(f"EVO_TTS_PROVIDER is set to '{preferred}', which is not one of: {', '.join(PROVIDERS)}")
+        if has_gemini_credentials():
+            return "gemini"
         if has_vbee_credentials():
             return "vbee"
         if has_openai_credentials():
             return "openai"
         raise TtsError(
-            "no TTS credentials found. Store Vbee with "
+            "no TTS credentials found. Store Gemini with "
+            "`evo cred add gemini_api_key --from-stdin`, Vbee with "
             "`evo cred add vbee.app_id --from-stdin` + `evo cred add vbee.token --from-stdin`, "
             "or OpenAI with `evo cred add openai_api_key --from-stdin`."
         )
@@ -42,14 +46,25 @@ def default_voice_for(provider):
 
 
 def default_voice(provider):
+    if provider == "gemini":
+        return gemini_tts.DEFAULT_VOICE
     return vbee.DEFAULT_VOICE if provider == "vbee" else openai_tts.DEFAULT_VOICE
 
 
 def supported_formats(provider):
+    if provider == "gemini":
+        return gemini_tts.FORMATS
     return vbee.FORMATS if provider == "vbee" else openai_tts.FORMATS
 
 
+def default_format(provider):
+    # Gemini only ever returns raw PCM, so wav is the format that needs no extra tooling.
+    return gemini_tts.DEFAULT_FORMAT if provider == "gemini" else "mp3"
+
+
 def chunk_limit(provider, mode):
+    if provider == "gemini":
+        return gemini_tts.TEXT_LIMIT
     if provider == "vbee":
         return vbee.BATCH_LIMIT if mode == "batch" else vbee.REALTIME_LIMIT
     return openai_tts.TEXT_LIMIT
@@ -136,6 +151,18 @@ def synthesize(
                 sample_rate=sample_rate,
             )
 
+    elif provider == "gemini":
+
+        def call(chunk):
+            return gemini_tts.synthesize(
+                chunk,
+                voice=voice,
+                output_format=output_format,
+                model=model or gemini_tts.DEFAULT_MODEL,
+                instructions=instructions,
+                bitrate=bitrate,
+            )
+
     else:
 
         def call(chunk):
@@ -198,6 +225,8 @@ def synthesize_many(items, concurrency=4, on_item=None, **kwargs):
 
 def list_voices(provider="auto", language_code=None, gender=None, ownership="VBEE", limit=100):
     provider = resolve_provider(provider)
+    if provider == "gemini":
+        return gemini_tts.list_voices()
     if provider == "openai":
         return openai_tts.list_voices()
     voices, pagination = vbee.list_voices(

{evo_cli-0.21.2 → evo_cli-0.22.0}/evo_cli/tts/creds.py

@@ -55,5 +55,24 @@ def has_openai_credentials():
     return True
 
 
+def gemini_api_key():
+    key = _resolve("GEMINI_API_KEY", "gemini_api_key") or os.environ.get("GOOGLE_API_KEY")
+    if not key:
+        raise TtsError(
+            "missing Gemini credentials: gemini_api_key\n"
+            "Get one at https://aistudio.google.com/apikey, then store it with:\n"
+            "  evo cred add gemini_api_key --from-stdin"
+        )
+    return key
+
+
+def has_gemini_credentials():
+    try:
+        gemini_api_key()
+    except TtsError:
+        return False
+    return True
+
+
 def vbee_webhook_url():
     return _resolve("VBEE_WEBHOOK_URL", "vbee.webhook_url")

{evo_cli-0.21.2 → evo_cli-0.22.0}/evo_cli/tts/errors.py

@@ -1,2 +1,2 @@
 class TtsError(RuntimeError):
-    pass
+    status = None