livekit-plugins-duplexjev 0.3.0__tar.gz → 0.3.2__tar.gz

{livekit_plugins_duplexjev-0.3.0 → livekit_plugins_duplexjev-0.3.2}/CHANGELOG.md

@@ -1,5 +1,17 @@
 # Changelog
 
+## 0.3.2 (2026-10-13)
+
+- Sends an `X-DuplexJev-Client: livekit-plugins-duplexjev/<version>` header (no personal data), so trial-API usage can be counted per integration.
+
+## 0.3.1 (2026-10-13)
+
+- Logs a one-time hint when the free trial API is used: it runs DuplexJev-4B-Para, so
+  its turn-taking does not match the published eot-bench numbers (DuplexJev-32B-Para v1.2 behind your own
+  `duplexjev gateway` >= 0.6).
+- README: how to reproduce the eot-bench numbers; the self-hosting example now serves DuplexJev-32B-Para v1.2 with
+  `--rate 1000000`; eot-bench figure for the default setting updated to the full run (20.0%).
+
 ## 0.3.0 (2026-10-13)
 
 - `readout="act"`: ask "What should the assistant do now?" and report P(reply now) + P(short acknowledgement) as the

{livekit_plugins_duplexjev-0.3.0 → livekit_plugins_duplexjev-0.3.2}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.5
 Name: livekit-plugins-duplexjev
-Version: 0.3.0
+Version: 0.3.2
 Summary: Audio end-of-turn detection for LiveKit Agents with DuplexJev: turn state, emotion and non-verbal sounds from one request, no transcript needed.
 Project-URL: Homepage, https://github.com/adventists-ai/duplexjev/tree/main/integrations/livekit
 Project-URL: Models, https://huggingface.co/adventists-ai/DuplexJev-4B-Para
@@ -43,6 +43,12 @@ say, and tells your agent whether the user is **finished**, **still talking**, *
 
 Tested with `livekit-agents` 1.8.6. Maintained by Adventists.ai (AI降临派).
 
+> **Getting the published turn-taking numbers.** LiveKit eot-bench (English, false cut-offs at 300 ms): **20.0%** for
+> DuplexJev-32B-Para v1.2 with the agent's previous message, 23.5% audio only, vs 35.2% for Smart Turn v3.2. That needs
+> DuplexJev-32B-Para **v1.2 on your own GPU** behind `duplexjev gateway` >= 0.6 (see *Run the model yourself*), the
+> default `readout="turn"`, and the agent's previous message (below). The free trial API runs the smaller DuplexJev-4B-Para
+> (no v1.2 turn-taking update): use it to try the plugin, not to judge turn-taking quality.
+
 ## Install
 
 ```bash
@@ -75,8 +81,8 @@ session = AgentSession(..., turn_handling={"turn_detection": detector})
 detector.attach(session)        # follows conversation_item_added; or detector.set_assistant_said(text)
 ```
 
-It needs a gateway from `duplexjev>=0.6`. On LiveKit eot-bench (English, DuplexJev-32B-Para v1.2) this cuts false
-cut-offs at a 300 ms budget from 23.5% to 19.7%.
+It needs a gateway from `duplexjev>=0.6`. On LiveKit eot-bench (English, DuplexJev-32B-Para v1.2, full run) this cuts
+false cut-offs at a 300 ms budget from 23.5% to 20.0%.
 
 `AgentSession(turn_detection=duplexjev.TurnDetector(), ...)` works too (LiveKit marks that argument as deprecated).
 A complete agent is in [`examples/agent.py`](examples/agent.py).
@@ -92,14 +98,18 @@ commits the turn, as with any turn detector, because the LLM needs the text.
 
 ## Run the model yourself (recommended for real-time)
 
-The default URL is our free trial API (rate limited to 30 requests per minute; the round trip from outside mainland China adds network latency). For a live agent, serve the model next to it (DuplexJev-4B-Para needs about 10 GB of GPU memory; for the best turn-taking use DuplexJev-32B-Para v1.2 on one 80 GB GPU):
+The default URL is our free trial API (rate limited to 30 requests per minute; the round trip from outside mainland China adds network latency; it runs the smaller DuplexJev-4B-Para without the v1.2 turn-taking update). For a live agent, serve the model next to it. For turn-taking, use DuplexJev-32B-Para v1.2 on one 80 GB GPU (the eot-bench numbers above):
 
 ```bash
 pip install "vllm[audio]>=0.29" duplexjev-vllm "duplexjev[server]>=0.6"
-vllm serve adventists-ai/DuplexJev-4B-Para --max-model-len 4096 --port 8000 &
-duplexjev gateway --vllm http://localhost:8000/v1 --port 8420
+vllm serve adventists-ai/DuplexJev-32B-Para --revision v1.2 --max-model-len 4096 --port 8000 &
+duplexjev gateway --vllm http://localhost:8000/v1 --port 8420 --rate 1000000
 ```
 
+On a smaller GPU, `adventists-ai/DuplexJev-4B-Para` needs about 10 GB; it is fine for emotion and the other questions,
+but its turn-taking has not had the v1.2 update. `--rate 1000000` lifts the gateway's default limit of 30 requests per
+minute, which a live agent exceeds quickly.
+
 Then pass `api_url="http://localhost:8420"` (or set `DUPLEXJEV_API_URL`).
 
 ## Options

{livekit_plugins_duplexjev-0.3.0 → livekit_plugins_duplexjev-0.3.2}/README.md

@@ -17,6 +17,12 @@ say, and tells your agent whether the user is **finished**, **still talking**, *
 
 Tested with `livekit-agents` 1.8.6. Maintained by Adventists.ai (AI降临派).
 
+> **Getting the published turn-taking numbers.** LiveKit eot-bench (English, false cut-offs at 300 ms): **20.0%** for
+> DuplexJev-32B-Para v1.2 with the agent's previous message, 23.5% audio only, vs 35.2% for Smart Turn v3.2. That needs
+> DuplexJev-32B-Para **v1.2 on your own GPU** behind `duplexjev gateway` >= 0.6 (see *Run the model yourself*), the
+> default `readout="turn"`, and the agent's previous message (below). The free trial API runs the smaller DuplexJev-4B-Para
+> (no v1.2 turn-taking update): use it to try the plugin, not to judge turn-taking quality.
+
 ## Install
 
 ```bash
@@ -49,8 +55,8 @@ session = AgentSession(..., turn_handling={"turn_detection": detector})
 detector.attach(session)        # follows conversation_item_added; or detector.set_assistant_said(text)
 ```
 
-It needs a gateway from `duplexjev>=0.6`. On LiveKit eot-bench (English, DuplexJev-32B-Para v1.2) this cuts false
-cut-offs at a 300 ms budget from 23.5% to 19.7%.
+It needs a gateway from `duplexjev>=0.6`. On LiveKit eot-bench (English, DuplexJev-32B-Para v1.2, full run) this cuts
+false cut-offs at a 300 ms budget from 23.5% to 20.0%.
 
 `AgentSession(turn_detection=duplexjev.TurnDetector(), ...)` works too (LiveKit marks that argument as deprecated).
 A complete agent is in [`examples/agent.py`](examples/agent.py).
@@ -66,14 +72,18 @@ commits the turn, as with any turn detector, because the LLM needs the text.
 
 ## Run the model yourself (recommended for real-time)
 
-The default URL is our free trial API (rate limited to 30 requests per minute; the round trip from outside mainland China adds network latency). For a live agent, serve the model next to it (DuplexJev-4B-Para needs about 10 GB of GPU memory; for the best turn-taking use DuplexJev-32B-Para v1.2 on one 80 GB GPU):
+The default URL is our free trial API (rate limited to 30 requests per minute; the round trip from outside mainland China adds network latency; it runs the smaller DuplexJev-4B-Para without the v1.2 turn-taking update). For a live agent, serve the model next to it. For turn-taking, use DuplexJev-32B-Para v1.2 on one 80 GB GPU (the eot-bench numbers above):
 
 ```bash
 pip install "vllm[audio]>=0.29" duplexjev-vllm "duplexjev[server]>=0.6"
-vllm serve adventists-ai/DuplexJev-4B-Para --max-model-len 4096 --port 8000 &
-duplexjev gateway --vllm http://localhost:8000/v1 --port 8420
+vllm serve adventists-ai/DuplexJev-32B-Para --revision v1.2 --max-model-len 4096 --port 8000 &
+duplexjev gateway --vllm http://localhost:8000/v1 --port 8420 --rate 1000000
 ```
 
+On a smaller GPU, `adventists-ai/DuplexJev-4B-Para` needs about 10 GB; it is fine for emotion and the other questions,
+but its turn-taking has not had the v1.2 update. `--rate 1000000` lifts the gateway's default limit of 30 requests per
+minute, which a live agent exceeds quickly.
+
 Then pass `api_url="http://localhost:8420"` (or set `DUPLEXJEV_API_URL`).
 
 ## Options

{livekit_plugins_duplexjev-0.3.0 → livekit_plugins_duplexjev-0.3.2}/livekit/plugins/duplexjev/turn_detector.py

@@ -122,7 +122,8 @@ class TurnDetector:
         on_answers: optional callback ``f(answers: dict)`` called after every successful prediction.
         readout: ``"turn"`` (turn-state question, default) or ``"act"``: ask "What should the assistant do now?" and
             report P(reply now) + P(short acknowledgement); slightly better on eot-bench with the agent's previous
-            message (18.4% vs 19.7% false cut-offs at 300 ms, English, 32B-Para v1.2).
+            message (18.4% vs 19.7% false cut-offs at 300 ms, English, 32B-Para v1.2, fast mode; the default
+            ``"turn"`` scores 20.0% in a full eot-bench run).
         window_s: seconds of the current user turn sent with each request (the most recent ones).
         trim_silence: drop the quiet audio before the user started talking, keeping 0.3 s of pre-roll.
         prediction_timeout: how long LiveKit waits for a prediction before committing without one. Defaults to
@@ -157,6 +158,8 @@ class TurnDetector:
         self._decide_url = self._base_url + "/v1/decide"
         api_key = api_key or os.environ.get("DUPLEXJEV_API_KEY")
         self._headers = {"Authorization": f"Bearer {api_key}"} if api_key else {}
+        from .version import __version__ as _v
+        self._headers["X-DuplexJev-Client"] = f"livekit-plugins-duplexjev/{_v}"  # usage stats on the trial API
         self._lang = lang
         self._unlikely_threshold = unlikely_threshold
         self._backchannel_threshold = backchannel_threshold
@@ -168,6 +171,10 @@ class TurnDetector:
                 HOSTED_PREDICTION_TIMEOUT if self._base_url == DEFAULT_URL else SELF_HOSTED_PREDICTION_TIMEOUT
             )
         self._prediction_timeout = prediction_timeout
+        if self._base_url == DEFAULT_URL:
+            logger.info("Using the free DuplexJev trial API (DuplexJev-4B-Para, rate limited). "
+                        "For the turn-taking quality reported on eot-bench, serve DuplexJev-32B-Para v1.2 behind "
+                        "`duplexjev gateway` >= 0.6 and pass its URL; see the README.")
         self._fallback_probability = fallback_probability
         self._http_session = http_session
 

livekit_plugins_duplexjev-0.3.2/livekit/plugins/duplexjev/version.py

@@ -0,0 +1 @@
+__version__ = "0.3.2"

livekit_plugins_duplexjev-0.3.0/livekit/plugins/duplexjev/version.py

@@ -1 +0,0 @@
-__version__ = "0.3.0"