--- a/OllamaInterface/OllamaClient.py Thu Aug 08 18:33:49 2024 +0200
+++ b/OllamaInterface/OllamaClient.py Sun Aug 25 19:44:24 2024 +0200
@@ -45,11 +45,6 @@
from the 'ollama' server was received
@signal modelsList(modelNames:list[str]) emitted after the list of model
names was obtained from the 'ollama' server
- @signal detailedModelsList(models:list[dict]) emitted after the list of
- models was obtained from the 'ollama' server giving some model details
- @signal runningModelsList(models:list[dict]) emitted after the list of
- running models was obtained from the 'ollama' server giving some model
- execution details
@signal pullStatus(msg:str, id:str, total:int, completed:int) emitted to indicate
the status of a pull request as reported by the 'ollama' server
@signal serverVersion(version:str) emitted after the server version was obtained
@@ -63,8 +58,6 @@
replyReceived = pyqtSignal(str, str, bool)
modelsList = pyqtSignal(list)
- detailedModelsList = pyqtSignal(list)
- runningModelsList = pyqtSignal(list)
pullStatus = pyqtSignal(str, str, int, int)
serverVersion = pyqtSignal(str)
finished = pyqtSignal()
@@ -247,79 +240,65 @@
Public method to request a list of models available locally from the 'ollama'
server with some model details.
"""
- # TODO: not implemented yet
- self.__sendRequest("tags", processResponse=self.__processDetailedModelsList)
+ response = self.__sendSyncRequest("tags")
- def __processDetailedModelsList(self, response):
- """
- Private method to process the tags response of the 'ollama' server extracting
- some model details.
-
- @param response dictionary containing the tags response
- @type dict
- """
models = []
- with contextlib.suppress(KeyError):
- for model in response["models"]:
- name = model["name"]
- if name:
- models.append(
- {
- "name": name,
- "id": model["digest"][:20], # first 20 characters only
- "size": model["size"],
- "modified": datetime.datetime.fromisoformat(
- model["modified_at"]
- ),
- }
- )
- self.detailedModelsList.emit(models)
+ if response is not None:
+ with contextlib.suppress(KeyError):
+ for model in response["models"]:
+ name = model["name"]
+ if name:
+ models.append(
+ {
+ "name": name,
+ "id": model["digest"][:20], # first 20 characters only
+ "size": model["size"],
+ "modified": datetime.datetime.fromisoformat(
+ model["modified_at"]
+ ),
+ }
+ )
+
+ return models
def listRunning(self):
"""
Public method to request a list of running models from the 'ollama' server.
"""
- # TODO: not implemented yet
- self.__sendRequest("ps", processResponse=self.__processRunningModelsList)
+ response = self.__sendSyncRequest("ps")
- def __processRunningModelsList(self, response):
- """
- Private method to process the ps response of the 'ollama' server extracting
- some model execution details.
-
- @param response dictionary containing the ps response
- @type dict
- """
models = []
- with contextlib.suppress(KeyError):
- for model in response["models"]:
- name = model["name"]
- if name:
- if model["size_vram"] == 0:
- processor = self.tr("100% CPU")
- elif model["size_vram"] == model["size"]:
- processor = self.tr("100% GPU")
- elif model["size_vram"] > model["size_"] or model["size"] == 0:
- processor = self.tr("unknown")
- else:
- sizeCpu = model["size"] - model["size_vram"]
- cpuPercent = round(sizeCpu / model["size_vram"] * 100)
- processor = self.tr("{0}% / {1}% CPU / GPU").format(
- cpuPercent, 100 - cpuPercent
+ if response is not None:
+ with contextlib.suppress(KeyError):
+ for model in response["models"]:
+ name = model["name"]
+ if name:
+ if model["size_vram"] == 0:
+ processor = self.tr("100% CPU")
+ elif model["size_vram"] == model["size"]:
+ processor = self.tr("100% GPU")
+ elif model["size_vram"] > model["size"] or model["size"] == 0:
+ processor = self.tr("unknown")
+ else:
+ sizeCpu = model["size"] - model["size_vram"]
+ cpuPercent = round(sizeCpu / model["size_vram"] * 100)
+ processor = self.tr("{0}% / {1}% CPU / GPU").format(
+ cpuPercent, 100 - cpuPercent
+ )
+ models.append(
+ {
+ "name": name,
+ "id": model["digest"][:20], # first 20 characters only
+ "size": model["size"],
+ "size_vram": model["size_vram"],
+ "processor": processor,
+ "expires": datetime.datetime.fromisoformat(
+ model["expires_at"]
+ ),
+ }
)
- models.append(
- {
- "name": name,
- "id": model["digest"][:20], # first 20 characters only
- "size": model["size"],
- "size_vram": model["size_vram"],
- "processor": processor,
- "expires": datetime.datetime.fromisoformat(
- model["expires_at"]
- ),
- }
- )
- self.runningModelsList.emit(models)
+
+ return models
def version(self):
"""
@@ -346,22 +325,20 @@
"""
return self.__state
- def __sendRequest(self, endpoint, data=None, processResponse=None):
+ def __getServerReply(self, endpoint, data=None):
"""
- Private method to send a request to the 'ollama' server and handle its
- responses.
+ Private method to send a request to the 'ollama' server and return a reply
+ object.
@param endpoint 'ollama' API endpoint to be contacted
@type str
@param data dictionary containing the data to send to the server
(defaults to None)
@type dict (optional)
- @param processResponse function handling the received data (defaults to None)
- @type function (optional)
+ @return 'ollama' server reply
+ @rtype QNetworkReply
"""
- self.__state = OllamaClientState.Requesting
-
- ollamaUrl = QUrl(
+ ollamaUrl = QUrl(
"{0}://{1}:{2}/api/{3}".format(
self.__plugin.getPreferences("OllamaScheme"),
(
@@ -386,9 +363,53 @@
reply = self.__networkManager.post(request, jsonData)
else:
reply = self.__networkManager.get(request)
+ reply.errorOccurred.connect(lambda error: self.__errorOccurred(error, reply))
+ return reply
+ def __sendRequest(self, endpoint, data=None, processResponse=None):
+ """
+ Private method to send a request to the 'ollama' server and handle its
+ responses.
+
+ @param endpoint 'ollama' API endpoint to be contacted
+ @type str
+ @param data dictionary containing the data to send to the server
+ (defaults to None)
+ @type dict (optional)
+ @param processResponse function handling the received data (defaults to None)
+ @type function (optional)
+ """
+ self.__state = OllamaClientState.Requesting
+
+ ##ollamaUrl = QUrl(
+ ##"{0}://{1}:{2}/api/{3}".format(
+ ##self.__plugin.getPreferences("OllamaScheme"),
+ ##(
+ ##"127.0.0.1"
+ ##if self.__localServer
+ ##else self.__plugin.getPreferences("OllamaHost")
+ ##),
+ ##(
+ ##self.__plugin.getPreferences("OllamaLocalPort")
+ ##if self.__localServer
+ ##else self.__plugin.getPreferences("OllamaPort")
+ ##),
+ ##endpoint,
+ ##)
+ ##)
+ ##request = QNetworkRequest(ollamaUrl)
+ ##if data is not None:
+ ##request.setHeader(
+ ##QNetworkRequest.KnownHeaders.ContentTypeHeader, "application/json"
+ ##)
+ ##jsonData = json.dumps(data).encode("utf-8")
+ ##reply = self.__networkManager.post(request, jsonData)
+ ##else:
+ ##reply = self.__networkManager.get(request)
+##
+ reply = self.__getServerReply(endpoint=endpoint, data=data)
reply.finished.connect(lambda: self.__replyFinished(reply))
- reply.errorOccurred.connect(lambda error: self.__errorOccurred(error, reply))
+ ##reply.errorOccurred.connect(lambda error: self.__errorOccurred(error, reply))
reply.readyRead.connect(lambda: self.__processData(reply, processResponse))
self.__replies.append(reply)
@@ -441,6 +462,36 @@
if data and processResponse:
processResponse(data)
+ def __sendSyncRequest(self, endpoint, data=None):
+ """
+ Private method to send a request to the 'ollama' server and handle its
+ responses.
+
+ @param endpoint 'ollama' API endpoint to be contacted
+ @type str
+ @param data dictionary containing the data to send to the server
+ (defaults to None)
+ @type dict (optional)
+ """
+ self.__state = OllamaClientState.Requesting
+
+ reply = self.__getServerReply(endpoint=endpoint, data=data)
+ while not reply.isFinished():
+ QCoreApplication.processEvents()
+ QThread.msleep(100)
+
+ reply.deleteLater()
+
+ self.__state = OllamaClientState.Finished
+
+ if reply.error() == QNetworkReply.NetworkError.NoError:
+ buffer = bytes(reply.readAll())
+ with contextlib.suppress(json.JSONDecodeError):
+ data = json.loads(buffer)
+ return data
+
+ return None
+
def heartbeat(self):
"""
Public method to check, if the 'ollama' server has started and is responsive.
