forked from SciSharp/BotSharp
-
Notifications
You must be signed in to change notification settings - Fork 0
Expand file tree
/
Copy pathSequentialPlanner.cs
More file actions
219 lines (181 loc) · 7.89 KB
/
SequentialPlanner.cs
File metadata and controls
219 lines (181 loc) · 7.89 KB
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
/*****************************************************************************
Copyright 2024 Written by Haiping Chen. All Rights Reserved.
Licensed under the Apache License, Version 2.0 (the "License");
you may not use this file except in compliance with the License.
You may obtain a copy of the License at
http://www.apache.org/licenses/LICENSE-2.0
Unless required by applicable law or agreed to in writing, software
distributed under the License is distributed on an "AS IS" BASIS,
WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
See the License for the specific language governing permissions and
limitations under the License.
******************************************************************************/
using BotSharp.Abstraction.Models;
using BotSharp.Abstraction.Settings;
namespace BotSharp.Plugin.Planner.Sequential;
/// <summary>
/// A Sequential Planner for a large language model (LLM) is a framework or methodology to execute some list tasks in a predefined order by the user,
/// LLM can follow to produce an organized and coherent output.
/// Sequential Planners are useful for tasks that involve multiple stages of reasoning, information retrieval, or interdependent subtasks.
/// </summary>
public class SequentialPlanner : ITaskPlanner
{
private readonly IServiceProvider _services;
private readonly ILogger _logger;
public string Name => "Sequential-Planner";
public bool HideDialogContext => true;
public int MaxLoopCount => 100;
private FunctionCallFromLlm _lastInst;
public SequentialPlanner(IServiceProvider services, ILogger<SequentialPlanner> logger)
{
_services = services;
_logger = logger;
}
public async Task<FunctionCallFromLlm> GetNextInstruction(Agent router, string messageId, List<RoleDialogModel> dialogs)
{
var decomposation = await GetDecomposedStepAsync(router, messageId, dialogs);
if (decomposation.TotalRemainingSteps > 0 && _lastInst != null)
{
// _lastInst.Response = decomposation.Description;
_lastInst.NextActionReason = $"Having {decomposation.TotalRemainingSteps} steps left.";
return _lastInst;
}
else if (decomposation.TotalRemainingSteps == 0 || decomposation.ShouldStop)
{
if (!string.IsNullOrEmpty(decomposation.StopReason))
{
// Tell router all steps are done
dialogs.Add(new RoleDialogModel(AgentRole.Assistant, decomposation.StopReason)
{
CurrentAgentId = router.Id,
MessageId = messageId
});
router.TemplateDict["conversation"] = router.TemplateDict["conversation"].ToString().TrimEnd() +
$"\r\n{router.Name}: {decomposation.StopReason}";
}
}
var next = GetNextStepPrompt(router);
// text completion
/*var agentService = _services.GetRequiredService<IAgentService>();
var instruction = agentService.RenderedInstruction(router);
var content = $"{instruction}\r\n###\r\n{next}";
content = content + "\r\nResponse: ";
var completion = CompletionProvider.GetTextCompletion(_services);*/
// chat completion
var completion = CompletionProvider.GetChatCompletion(_services,
provider: router?.LlmConfig?.Provider,
model: router?.LlmConfig?.Model);
string text = string.Empty;
// text completion
// text = await completion.GetCompletion(content, router.Id, messageId);
dialogs = new List<RoleDialogModel>
{
new RoleDialogModel(AgentRole.User, next)
{
FunctionName = nameof(SequentialPlanner),
MessageId = messageId
}
};
var response = await completion.GetChatCompletions(router, dialogs);
var inst = response.Content.JsonContent<FunctionCallFromLlm>();
if (decomposation.TotalRemainingSteps > 0)
{
// inst.Response = decomposation.Description;
inst.NextActionReason = $"{decomposation.TotalRemainingSteps} steps left.";
inst.HandleDialogsByPlanner = true;
}
_lastInst = inst;
return inst;
}
public List<RoleDialogModel> BeforeHandleContext(FunctionCallFromLlm inst, RoleDialogModel message, List<RoleDialogModel> dialogs)
{
var taskAgentDialogs = new List<RoleDialogModel>
{
/*new RoleDialogModel(AgentRole.User, inst.Response)
{
MessageId = message.MessageId,
}*/
};
return taskAgentDialogs;
}
public bool AfterHandleContext(List<RoleDialogModel> dialogs, List<RoleDialogModel> taskAgentDialogs)
{
dialogs.AddRange(taskAgentDialogs.Skip(1));
return true;
}
public Task<bool> AgentExecuting(Agent router, FunctionCallFromLlm inst, RoleDialogModel message, List<RoleDialogModel> dialogs)
{
// Set user content as Planner's question
message.FunctionName = inst.Function;
message.FunctionArgs = inst.Arguments == null ? "{}" : JsonSerializer.Serialize(inst.Arguments);
return Task.FromResult(true);
}
public async Task<bool> AgentExecuted(Agent router, FunctionCallFromLlm inst, RoleDialogModel message, List<RoleDialogModel> dialogs)
{
var context = _services.GetRequiredService<IRoutingContext>();
if (message.StopCompletion)
{
await context.Empty(reason: $"Agent queue is cleared by {nameof(SequentialPlanner)}");
return false;
}
// Handover to Router;
await context.Pop();
var routing = _services.GetRequiredService<IRoutingService>();
routing.Context.ResetRecursiveCounter();
return true;
}
private string GetNextStepPrompt(Agent router)
{
var template = router.Templates.First(x => x.Name == "reasoner.sequential").Content;
var render = _services.GetRequiredService<ITemplateRender>();
return render.Render(template, new Dictionary<string, object>
{
});
}
public async Task<DecomposedStep> GetDecomposedStepAsync(Agent router, string messageId, List<RoleDialogModel> dialogs)
{
var systemPrompt = GetDecomposeTaskPrompt(router);
var inst = new DecomposedStep();
var llmProviderService = _services.GetRequiredService<ILlmProviderService>();
var settingService = _services.GetRequiredService<ISettingService>();
var model = llmProviderService.GetProviderModel("openai", settingService.GetUpgradeModel(Gpt4xModelConstants.GPT_4o));
// chat completion
var completion = CompletionProvider.GetChatCompletion(_services,
provider: "openai",
model: model?.Name);
int retryCount = 0;
while (retryCount < 2)
{
string text = string.Empty;
try
{
var response = await completion.GetChatCompletions(new Agent
{
Id = router.Id,
Name = nameof(SequentialPlanner),
Instruction = systemPrompt
}, dialogs);
text = response.Content;
inst = response.Content.JsonContent<DecomposedStep>();
break;
}
catch (Exception ex)
{
_logger.LogError($"{ex.Message}: {text}");
}
finally
{
retryCount++;
}
}
return inst;
}
private string GetDecomposeTaskPrompt(Agent router)
{
var template = router.Templates.First(x => x.Name == "reasoner.sequential.get_remaining_task").Content;
var render = _services.GetRequiredService<ITemplateRender>();
return render.Render(template, new Dictionary<string, object>
{
});
}
}