feat(regen): agent force-end-turn and Flux expressivity controls

.agents/skills/deepgram-python-conversational-stt/SKILL.md

@@ -82,23 +82,25 @@ with client.listen.v2.connect(
 | `encoding` | `linear16`, `mulaw`, etc. Omit for containerized audio |
 | `sample_rate` | String in the SDK signature, e.g. `"16000"` |
 | `eager_eot_threshold` | Fire end-of-turn early at this confidence |
-| `eot_threshold` | Primary end-of-turn confidence |
-| `eot_timeout_ms` | Time-based fallback turn end |
+| `eot_threshold` | Primary end-of-turn confidence; set to `"1.0"` to suppress confidence-based endings |
+| `eot_timeout_ms` | Time-based fallback turn end; still applies when `eot_threshold="1.0"` |
 | `keyterm` | Bias for domain keywords |
 | `mip_opt_out`, `tag` | Metadata / privacy flags |
 | `language_hint` | **ONLY for `flux-general-multi`** |
 | `authorization`, `request_options` | Override auth or request options |
 
 **No `language` parameter** on v2 — language is implied by model (`flux-general-en`) or hinted via `language_hint` on multi.
 
+For application-controlled turns, use `eot_threshold="1.0"` with a sufficiently large `eot_timeout_ms`, then call `conn.send_force_end_turn()` for the active turn. ForceEndTurn requires deployment enablement; see `examples/16-transcription-force-end-turn.py`.
+
 ## Events (server → client)
 
 - `ListenV2Connected` — connection established
 - `ListenV2ConfigureSuccess` / `ListenV2ConfigureFailure` — mid-session config changes
 - `ListenV2TurnInfo` — per-turn transcript + event (`Update`, `EndOfTurn`, `EagerEndOfTurn`, ...) + `turn_index`
 - `ListenV2FatalError` — terminal error
 
-Client messages: `ListenV2Media`, `ListenV2Configure`, `ListenV2CloseStream`.
+Client messages: `ListenV2Media`, `ListenV2Configure`, `ListenV2ForceEndTurn`, `ListenV2CloseStream`.
 
 ## Async equivalent
 

.agents/skills/deepgram-python-voice-agent/SKILL.md

@@ -119,6 +119,7 @@ with client.agent.v1.connect() as agent:
 - Prompt / think / speak update messages (change mid-session)
 - User / assistant text injection
 - Function call response (reply to `FunctionCallRequest`)
+- `ForceEndTurn` (end an active user turn; requires a Deepgram V2/Flux listen provider)
 
 ## Reusable agent configurations
 
@@ -190,13 +191,18 @@ agent.send_inject_user_message(
 # 6. Idle-period keep-alive (no payload required; the SDK fills in the type literal)
 agent.send_keep_alive(AgentV1KeepAlive())
 # Or simply: agent.send_keep_alive()  — the message arg is optional.
+
+# 7. End an active user turn immediately (for example, on push-to-talk release).
+# Requires a Deepgram V2/Flux listen provider; V1 returns FORCE_END_TURN_UNSUPPORTED.
+agent.send_force_end_turn()
 ```
 
 Async client equivalents are identical but `await`-prefixed:
 
 ```python
 await agent.send_update_prompt(AgentV1UpdatePrompt(prompt="..."))
 await agent.send_inject_agent_message(AgentV1InjectAgentMessage(message="..."))
+await agent.send_force_end_turn()
 ```
 
 ## Stream lifecycle & recovery
@@ -288,6 +294,7 @@ The server emits a `History` message on connect when the SDK has captured prior
 ## Example files in this repo
 
 - `examples/30-voice-agent.py`
+- `examples/32-voice-agent-force-end-turn.py` — Force an active turn to end with a Flux listen provider
 - `tests/manual/agent/v1/connect/main.py` — live connection test
 
 ## Central product skills

.fern/metadata.json

@@ -21,8 +21,8 @@
       "skip_validation": true
     }
   },
-  "originGitCommit": "068de888501fcc3b792086aab45de975587b89e1",
+  "originGitCommit": "0825695d67503a2d95eda2ecb88672d7ee6aa4d6",
   "originGitCommitIsDirty": true,
   "invokedBy": "manual",
-  "sdkVersion": "7.8.1"
+  "sdkVersion": "7.8.2"
 }

.fernignore

@@ -132,15 +132,6 @@ src/deepgram/agent/v1/types/agent_v1settings_agent_context_listen_provider.py
 # when the old provider payloads are retired in a future major.
 src/deepgram/agent/v1/types/agent_v1update_listen_listen.py
 
-# Backward-compat patch for the 2026-08-18 regen SpeakV2Speed retype. The generator
-# changed this from `float` to `Union[Literal["0.85"..."1.15"], Any]` -- a string-literal
-# enum that contradicts the API contract (a numeric multiplier; cf. SpeakV2SpeedValue =
-# float, used by the Configure message) and silently changed the documented domain of the
-# `speak.v2.connect(speed=...)` param from numeric to string. Restored to `float` so the
-# connect param stays exactly what it was on main and consistent with Configure. Unfreeze
-# when the spec types the connect `speed` as a number.
-src/deepgram/types/speak_v2speed.py
-
 # Hand-written compat shim recreating ListenV2CloseStreamType, which Fern removed in the
 # 2026-06-15 regen (docs #946). The original generated type wrongly allowed
 # Union[Literal["Finalize","CloseStream","KeepAlive"], Any] — v2 copied v1's control-message

AGENTS.md

@@ -70,7 +70,6 @@ Current temporarily frozen files:
 - `src/deepgram/core/query_encoder.py` — coerces Python bools to lowercase `"true"`/`"false"` before they reach `urllib.parse.urlencode` (which would otherwise produce `"True"`/`"False"` via `str()` and break websocket query strings). Only the four `*/connect()` paths call `urlencode`; HTTP raw clients hand params to httpx, which lowercases bools itself, so the patch is a no-op for the HTTP path. Once Fern's websocket codegen normalizes bools (or the spec types these as `boolean` end-to-end), this can be unfrozen.
 - `src/deepgram/listen/v2/types/listen_v2connected.py`, `src/deepgram/listen/v2/types/listen_v2turn_info.py`, `src/deepgram/listen/v2/types/listen_v2turn_info_words_item.py`, `src/deepgram/listen/v2/types/listen_v2configure_success.py`, `src/deepgram/listen/v2/types/listen_v2configure_success_thresholds.py`, `src/deepgram/listen/v2/types/listen_v2configure_failure.py`, `src/deepgram/listen/v2/types/listen_v2fatal_error.py` — read-side compatibility for the SDK 7.7 Listen V2 response retype. Through 7.6, `V2SocketClientResponse` contained `typing.Any`, so every response was returned as a raw dict; fixing the union made responses typed models and broke callers using `response["field"]`. These generated response classes inherit the hand-written base above, preserving read-only wire-key subscript access alongside canonical attribute access. Restore direct `UncheckedBaseModel` inheritance and unfreeze these files in the next major release.
 - `src/deepgram/types/deepgram_listen_provider_v2.py`, `src/deepgram/agent/v1/types/agent_v1settings_agent_listen_provider.py`, `src/deepgram/agent/v1/types/agent_v1settings_agent_context_listen_provider.py` — behavioural back-compat shim for the `language_hint` -> `language_hints` rename (2026-06-15 regen). The public field was historically (incorrectly) singular and accepted a str or a list; the API field is `language_hints` (a list, and the server uses `deny_unknown_fields` so the singular key is rejected on the wire). Each carries a hand-added `model_validator(mode='before')` / `root_validator(pre=True)` that remaps a legacy `language_hint=` kwarg and drops the dead singular key. Remove and unfreeze when the singular alias is retired in a future major.
-- `src/deepgram/types/speak_v2speed.py` — behavioural back-compat patch for the 2026-08-18 regen `SpeakV2Speed` retype. The generator changed it from `float` to `Union[Literal["0.85"…"1.15"], Any]` (a string-literal enum), which silently changed the documented domain of the `speak.v2.connect(speed=...)` parameter from numeric to string and contradicts the actual API contract (`SpeakV2SpeedValue = float`, used by the mid-stream `SpeakV2Configure` message). Only the `Any` fallback kept `speed=1.05` working. Restored to `float` so the connect param stays exactly what it was on `main`, mypy-precise, and consistent with the `Configure` message. Regression coverage in `tests/custom/test_speak_v2_coverage.py`. Unfreeze when the spec types the connect `speed` as a number.
 - `src/deepgram/agent/v1/types/agent_v1update_listen_listen.py` — backward-compat patch for the 2026-07-31 `AgentV1UpdateListen` provider retype. The `provider` field changed from a bare `DeepgramListenProviderV2` to the required discriminated union `AgentV1UpdateListenListenProvider` (`_V1`/`_V2`, discriminant `version`). Carries a hand-added `model_validator(mode='before')` / `root_validator(pre=True)` that coerces a legacy `DeepgramListenProviderV1`/`V2` (or a dict lacking the `version` discriminant) into the new shape so existing callers keep working. Remove and unfreeze when the old provider payloads are retired in a future major. NOTE: this patch was silently lost once (it was absent from `.fernignore`, so a regen overwrote it) — keep it frozen.
 - `tests/wire/test_manage_v1_projects_keys.py` — restored wire coverage for the legacy `CreateKeyV1RequestOneParams` request alias so future regens do not silently drop that compatibility check
 - `tests/wire/test_manage_v1_projects_requests.py` — restored query-parameter coverage for `manage.v1.projects.requests.list`. The 2026-08-11 regen simplified the upstream spec *example*, and Fern derives the wire test from the example, so all ten optional query params (and the `datetime` → ISO-8601 `Z` encoding) lost their assertions while the client signature still forwarded them. Frozen for the same reason as the `_keys.py` entry above.
@@ -87,7 +86,7 @@ Files Fern now owns outright, but that carry a caveat worth knowing before the n
 
 ### Prepare repo for regeneration
 
-1. **Create a new branch** off `main` named `lo/sdk-gen-<YYYY-MM-DD>`.
+1. **Create a new branch** off `main` named `gh/sdk-gen-<YYYY-MM-DD>`.
 2. **Push the branch** and create a PR titled `chore: SDK regeneration <YYYY-MM-DD>` (empty commit if needed).
 3. **Read `.fernignore`** and classify each entry using the rules above.
 4. **For each temporarily frozen file only:**

examples/32-voice-agent-force-end-turn.py

@@ -0,0 +1,96 @@
+"""
+Example: Force an active Voice Agent turn to end.
+
+ForceEndTurn requires a Deepgram V2 (Flux) listen provider. It sends the
+control message after the server reports UserStartedSpeaking, then waits for
+the agent response. The example streams the first two seconds of the bundled
+WAV fixture and does not capture microphone audio.
+"""
+
+import threading
+import time
+from pathlib import Path
+
+from dotenv import load_dotenv
+
+from deepgram import DeepgramClient
+from deepgram.agent.v1.types import (
+    AgentV1Settings,
+    AgentV1SettingsAgent,
+    AgentV1SettingsAgentListen,
+    AgentV1SettingsAgentListenProvider_V2,
+    AgentV1SettingsAudio,
+    AgentV1SettingsAudioInput,
+)
+from deepgram.core.events import EventType
+from deepgram.types.speak_settings_v1 import SpeakSettingsV1
+from deepgram.types.speak_settings_v1provider import SpeakSettingsV1Provider_Deepgram
+from deepgram.types.think_settings_v1 import ThinkSettingsV1
+from deepgram.types.think_settings_v1provider import ThinkSettingsV1Provider_OpenAi
+
+load_dotenv()
+
+AUDIO_PATH = Path(__file__).parent / "fixtures" / "audio.wav"
+
+
+def main() -> None:
+    user_started = threading.Event()
+    agent_finished = threading.Event()
+
+    settings = AgentV1Settings(
+        audio=AgentV1SettingsAudio(input=AgentV1SettingsAudioInput(encoding="linear16", sample_rate=44100)),
+        agent=AgentV1SettingsAgent(
+            listen=AgentV1SettingsAgentListen(
+                provider=AgentV1SettingsAgentListenProvider_V2(type="deepgram", model="flux-general-en")
+            ),
+            think=ThinkSettingsV1(
+                provider=ThinkSettingsV1Provider_OpenAi(type="open_ai", model="gpt-4o-mini"),
+                prompt="Reply briefly.",
+            ),
+            speak=SpeakSettingsV1(
+                provider=SpeakSettingsV1Provider_Deepgram(type="deepgram", model="aura-2-asteria-en")
+            ),
+        ),
+    )
+
+    with DeepgramClient().agent.v1.connect() as agent:
+
+        def on_message(message: object) -> None:
+            message_type = getattr(message, "type", None)
+            if message_type == "UserStartedSpeaking":
+                user_started.set()
+                print("UserStartedSpeaking received")
+            elif message_type == "ConversationText":
+                print(f"[{message.role}] {message.content}")
+            elif message_type == "AgentAudioDone":
+                agent_finished.set()
+                print("AgentAudioDone received")
+            elif message_type in {"Warning", "Error"}:
+                print(f"{message_type}: {message.code} - {message.description}")
+
+        agent.on(EventType.MESSAGE, on_message)
+        agent.on(EventType.ERROR, lambda error: print(f"Connection error: {error}"))
+        threading.Thread(target=agent.start_listening, daemon=True).start()
+
+        agent.send_settings(settings)
+        with AUDIO_PATH.open("rb") as audio_file:
+            audio_file.read(44)
+            audio = audio_file.read(44100 * 2 * 2)
+
+        for start in range(0, len(audio), 44100 // 10 * 2):
+            agent.send_media(audio[start : start + 44100 // 10 * 2])
+            time.sleep(0.1)
+            if user_started.is_set():
+                break
+
+        if not user_started.is_set():
+            raise TimeoutError("Timed out waiting for UserStartedSpeaking")
+
+        print("Sending ForceEndTurn")
+        agent.send_force_end_turn()
+        if not agent_finished.wait(15):
+            raise TimeoutError("Timed out waiting for the agent response")
+
+
+if __name__ == "__main__":
+    main()

examples/README.md

@@ -34,6 +34,7 @@ This directory contains comprehensive examples demonstrating how to use the Deep
 
 - **30-voice-agent.py** - Voice Agent configuration and usage
 - **31-voice-agent-session-recording.py** - Record selected Voice Agent events as JSON
+- **32-voice-agent-force-end-turn.py** - End an active Voice Agent Flux turn with ForceEndTurn
 
 ### 40-49: Text Intelligence (Read)