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
|
using System.Data;
using System.Text.Json;
using Tango.Portal.Chat.Web.Models;
using Tango.Portal.Chat.Web.Services;
using Kusto.Data.Data;
using Microsoft.AspNetCore.Mvc;
using Newtonsoft.Json.Linq;
using System.Text.Json.Nodes;
using System.Collections;
using Tango.Portal.Chat.Web.Utils;
namespace Tango.Portal.Chat.Web.Controllers
{
[ApiController]
[Route("api/[controller]")]
public sealed class ChatController : ControllerBase
{
private readonly SchemaRegistry _schema;
private readonly KqlGuard _guard;
private readonly KustoQueryService _adx;
private readonly LlmClient _llm;
private static readonly string[] AllowTables = new[] { "JobRunsTable", "JobStatusTable", "TelemetryTable", "MachinesTable" };
public ChatController(SchemaRegistry schema, KqlGuard guard, KustoQueryService adx, LlmClient llm)
{
_schema = schema;
_guard = guard;
_adx = adx;
_llm = llm;
}
[HttpPost("ask")]
public async Task<ActionResult<ChatResponse>> Ask([FromBody] ChatRequest req, CancellationToken ct)
{
try
{
var schemaJson = _schema.GetSchemaJson();
var plannerPrompt = _schema.GetPlannerPrompt();
var plotySample = _schema.GetPlotySample();
// 1) Ask the model for KQL
var plan = await _llm.ProposeKqlAsync(plannerPrompt, plotySample, req.Question, schemaJson, req.History, ct);
if (plan.Assistant == "data")
{
return await AnswerWithDataAssistant(req, plan, ct);
}
else if (plan.Assistant == "ploty")
{
return await AnswerWithDataAssistant(req, plan, ct);
}
else if (plan.Assistant == "docs")
{
return await AnswerWithDocsAssistant(req, plan, ct);
}
else
{
return AnswerWithPlannerConversation(req, plan);
}
}
catch (Exception ex)
{
return new ChatResponse
{
Answer = $"Ooops something went wrong...\n{ex.Message}",
ThreadId = req.ThreadId
};
}
}
private static ActionResult<ChatResponse> AnswerWithPlannerConversation(ChatRequest req, ProposeKqlResult plan)
{
return new ChatResponse
{
Answer = plan.ConversationAnswer,
ThreadId = req.ThreadId
};
}
private async Task<ActionResult<ChatResponse>> AnswerWithDocsAssistant(ChatRequest req, ProposeKqlResult plan, CancellationToken ct)
{
// AFTER
var run = await _llm.AnswerWithAssistantAsync(
LlmClient.AssistantType.Docs,
req.Question,
string.Empty,
plan.Kql,
req.ThreadId, // <-- reuse if provided
ct);
return new ChatResponse
{
Answer = run.Answer,
ThreadId = run.ThreadId
};
}
private async Task<ActionResult<ChatResponse>> AnswerWithDataAssistant(ChatRequest req, ProposeKqlResult plan, CancellationToken ct)
{
// 2) Guardrail validation
var val = _guard.Validate(plan.Kql);
if (!val.IsOk)
{
// Return error to the client so they can iterate
return new ChatResponse
{
Answer = $"The generated kusto query contains invalid tokens..\n{val.Error}",
ThreadId = req.ThreadId,
UsedKql = plan.Kql
};
}
// 4) Execute in ADX
DataTable table;
try
{
table = await _adx.QueryAsync(plan.Kql, plan.Parameters, ct);
}
catch (Exception ex)
{
// Return error to the client so they can iterate
return new ChatResponse
{
Answer = $"Seems like my kusto query ran into some issue..\n{ex.Message}",
ThreadId = req.ThreadId,
UsedKql = plan.Kql
};
}
if (table.Columns.Contains("ploty") && table.Rows.Count > 0)
{
return AnswerWithPloty(req, plan, table);
}
// 5) Build compact facts (limit rows/cols)
if (table.Rows.Count <= 200)
{
return await AnswerWithDataAssistantInternal(req, plan, table, ct);
}
else
{
return AnswerWithMarkdownTable(req, plan, table);
}
}
private async Task<ActionResult<ChatResponse>> AnswerWithDataAssistantInternal(ChatRequest req, ProposeKqlResult plan, DataTable table, CancellationToken ct)
{
var preview = DataHelper.ToPreview(table, 200);
var facts = JsonSerializer.Serialize(preview);
// 6) Ask model for final answer
//var answer = await _llm.AnswerFromFactsAsync(req.Question, facts, plan.Kql, ct);
var run = await _llm.AnswerWithAssistantAsync(
LlmClient.AssistantType.Data,
req.Question,
facts,
plan.Kql,
req.ThreadId, // <-- reuse if provided
ct);
return new ChatResponse
{
Answer = run.Answer,
UsedKql = plan.Kql,
Preview = preview,
ThreadId = run.ThreadId // <-- echo back the thread id used/created
};
}
private static ActionResult<ChatResponse> AnswerWithMarkdownTable(ChatRequest req, ProposeKqlResult plan, DataTable table)
{
var markdown = DataHelper.ToMarkdownTable(table);
return new ChatResponse
{
Answer = $"The result set was too big for me to analyze so I'm just going to drop the entire result set.\n{markdown}",
UsedKql = plan.Kql,
Preview = markdown,
ThreadId = req.ThreadId // <-- echo back the thread id used/created
};
}
private static ActionResult<ChatResponse> AnswerWithPloty(ChatRequest req, ProposeKqlResult plan, DataTable table)
{
String? ploty = table.Rows[0]["ploty"]?.ToString();
return new ChatResponse
{
Answer = plan.ConversationAnswer,
ThreadId = req.ThreadId,
UsedKql = plan.Kql,
Ploty = ploty ?? String.Empty
};
}
}
}
|
