--- a/OllamaInterface/OllamaClient.py Sun Aug 04 16:57:01 2024 +0200
+++ b/OllamaInterface/OllamaClient.py Mon Aug 05 18:37:16 2024 +0200
@@ -12,8 +12,16 @@
import enum
import json
-from PyQt6.QtCore import pyqtSignal, QObject, QUrl
-from PyQt6.QtNetwork import QNetworkAccessManager, QNetworkRequest, QNetworkReply
+from PyQt6.QtCore import (
+ QCoreApplication,
+ QObject,
+ QThread,
+ QTimer,
+ QUrl,
+ pyqtSignal,
+ pyqtSlot,
+)
+from PyQt6.QtNetwork import QNetworkAccessManager, QNetworkReply, QNetworkRequest
from eric7.EricNetwork.EricNetworkProxyFactory import proxyAuthenticationRequired
@@ -22,6 +30,7 @@
"""
Class defining the various client states.
"""
+
Waiting = 0
Requesting = 1
Receiving = 2
@@ -32,7 +41,7 @@
"""
Class implementing the 'ollama' client.
- @signal replyReceived(content:str, role:str) emitted after a response from the
+ @signal replyReceived(content:str, role:str) emitted after a response from the
'ollama' server was received
@signal modelsList(modelNames:list[str]) emitted after the list of model
names was obtained from the 'ollama' server
@@ -43,9 +52,13 @@
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
+ from the 'ollama' server
@signal finished() emitted to indicate the completion of a request
@signal errorOccurred(error:str) emitted to indicate a network error occurred
while processing the request
+ @signal serverStateChanged(ok:bool) emitted to indicate a change of the server
+ responsiveness
"""
replyReceived = pyqtSignal(str, str)
@@ -53,8 +66,10 @@
detailedModelsList = pyqtSignal(list)
runningModelsList = pyqtSignal(list)
pullStatus = pyqtSignal(str, str, int, int)
+ serverVersion = pyqtSignal(str)
finished = pyqtSignal()
errorOccurred = pyqtSignal(str)
+ serverStateChanged = pyqtSignal(bool)
def __init__(self, plugin, parent=None):
"""
@@ -75,8 +90,17 @@
proxyAuthenticationRequired
)
+ self.__serverResponding = False
+ self.__heartbeatTimer = QTimer(self)
+ self.__heartbeatTimer.timeout.connect(self.__periodicHeartbeat)
+
self.__state = OllamaClientState.Waiting
+ self.__serverResponding = False # start with a faulty state
+
+ self.__plugin.preferencesChanged.connect(self.__setHeartbeatTimer)
+ self.__setHeartbeatTimer()
+
def chat(self, model, messages):
"""
Public method to request a chat completion from the 'ollama' server.
@@ -188,7 +212,6 @@
Public method to request a list of models available locally from the 'ollama'
server.
"""
- # TODO: not implemented yet
self.__sendRequest("tags", processResponse=self.__processModelsList)
def __processModelsList(self, response):
@@ -203,16 +226,13 @@
for model in response["models"]:
name = model["name"]
if name:
- models.append(name)
+ models.append(name.replace(":latest", ""))
self.modelsList.emit(models)
def listDetails(self):
"""
Public method to request a list of models available locally from the 'ollama'
server with some model details.
-
- @return list of available models
- @rtype list of dict
"""
# TODO: not implemented yet
self.__sendRequest("tags", processResponse=self.__processDetailedModelsList)
@@ -244,7 +264,7 @@
def listRunning(self):
"""
- Public method to request a list of running models from the 'ollama' server
+ Public method to request a list of running models from the 'ollama' server.
"""
# TODO: not implemented yet
self.__sendRequest("ps", processResponse=self.__processRunningModelsList)
@@ -288,10 +308,26 @@
)
self.runningModelsList.emit(models)
+ def version(self):
+ """
+ Public method to request the version from the 'ollama' server.
+ """
+ self.__sendRequest("version", processResponse=self.__processVersion)
+
+ def __processVersion(self, response):
+ """
+ Private method to process the version response of the 'ollama' server.
+
+ @param response dictionary containing the version response
+ @type dict
+ """
+ with contextlib.suppress(KeyError):
+ self.serverVersion.emit(response["version"])
+
def state(self):
"""
Public method to get the current client state.
-
+
@return current client state
@rtype OllamaClientState
"""
@@ -326,9 +362,9 @@
QNetworkRequest.KnownHeaders.ContentTypeHeader, "application/json"
)
jsonData = json.dumps(data).encode("utf-8")
- reply = self.__networkManager.post(request=request, data=jsonData)
+ reply = self.__networkManager.post(request, jsonData)
else:
- reply = self.__networkManager.get(request=request)
+ reply = self.__networkManager.get(request)
reply.finished.connect(lambda: self.__replyFinished(reply))
reply.errorOccurred.connect(lambda error: self.__errorOccurred(error, reply))
@@ -383,3 +419,49 @@
data = json.loads(buffer)
if data and processResponse:
processResponse(data)
+
+ def heartbeat(self):
+ """
+ Public method to check, if the 'ollama' server has started and is responsive.
+
+ @return flag indicating a responsive 'ollama' server
+ @rtype bool
+ """
+ ollamaUrl = QUrl(
+ "{0}://{1}:{2}/".format(
+ self.__plugin.getPreferences("OllamaScheme"),
+ self.__plugin.getPreferences("OllamaHost"),
+ self.__plugin.getPreferences("OllamaPort"),
+ )
+ )
+ request = QNetworkRequest(ollamaUrl)
+ reply = self.__networkManager.head(request)
+ while not reply.isFinished():
+ QCoreApplication.processEvents()
+ QThread.msleep(100)
+
+ reply.deleteLater()
+
+ return reply.error() == QNetworkReply.NetworkError.NoError
+
+ @pyqtSlot()
+ def __setHeartbeatTimer(self):
+ """
+ Private slot to configure the heartbeat timer.
+ """
+ interval = self.__plugin.getPreferences("OllamaHeartbeatInterval")
+ if interval:
+ self.__heartbeatTimer.setInterval(interval * 1000) # interval in ms
+ self.__heartbeatTimer.start()
+ else:
+ self.__heartbeatTimer.stop()
+
+ @pyqtSlot()
+ def __periodicHeartbeat(self):
+ """
+ Private slot to do a periodic check of the 'ollama' server responsiveness.
+ """
+ responding = self.heartbeat()
+ if responding != self.__serverResponding:
+ self.serverStateChanged.emit(responding)
+ self.__serverResponding = responding
