File size: 10,494 Bytes
84983c9
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
a44477f
84983c9
 
a44477f
400e193
 
 
 
84983c9
 
 
 
 
 
 
 
 
 
 
 
 
 
 
a44477f
84983c9
 
 
 
 
 
 
 
 
 
 
 
400e193
84983c9
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
a44477f
84983c9
 
 
 
 
a44477f
84983c9
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
a44477f
84983c9
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
a44477f
84983c9
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
"""OpenAI tool-calling agent — catalog as external memory (no full DB in prompts)."""

from __future__ import annotations

import json
import os
import sqlite3
from typing import Any

from openai import OpenAI

from server.catalog import connect, get_commodity, search_catalog, summarize_row


def _parse_json_response(text: str) -> dict[str, Any] | None:
    t = (text or "").strip()
    if t.startswith("```"):
        lines = t.splitlines()
        if lines and lines[0].lstrip().startswith("```"):
            lines = lines[1:]
        if lines and lines[-1].strip() == "```":
            lines = lines[:-1]
        t = "\n".join(lines).strip()
    try:
        out = json.loads(t)
        return out if isinstance(out, dict) else None
    except json.JSONDecodeError:
        return None


def _assistant_message_dict(msg: Any) -> dict[str, Any]:
    d: dict[str, Any] = {"role": "assistant", "content": msg.content}
    if getattr(msg, "tool_calls", None):
        d["tool_calls"] = [
            {
                "id": tc.id,
                "type": "function",
                "function": {
                    "name": tc.function.name,
                    "arguments": tc.function.arguments or "{}",
                },
            }
            for tc in msg.tool_calls
        ]
    return d

SYSTEM = """You are ProcureMind, an AI procurement assistant with access to the procurement catalogue through tools only.

Rules:
- Never invent catalogue codes. Every code must come from tool results.
- Call `search_catalog` with **short procurement phrases** (about 2–6 keywords), not a pasted paragraph. Example: "consumer research surveys" or "market analysis studies" rather than the user's full sentence.
- **Retry before giving up:** If matches are empty or a poor fit, run additional `search_catalog` calls with broader or alternate terms (e.g. consumer research, marketing studies, survey services, statistical analysis, opinion polls). UNSPSC uses formal titles — paraphrase the user's intent into catalogue-style wording.
- If results are ambiguous but plausible, return status "choose" with 2–6 distinct candidates rather than "not_found".
- Only return status "not_found" after **at least two** distinct searches still yield nothing usable; then suggest narrower catalogue-style keywords.
- When you have a single best commodity, return status "found" with that commodity_code (integer).
- Always include human-readable code references: segment / family / class / commodity integers exactly as returned by tools.

Final reply MUST be a single JSON object (no markdown fences) with this shape:
{
  "status": "found" | "choose" | "not_found",
  "summary": "short natural language for the user",
  "commodity_code": null or integer,
  "candidates": [
     {"commodity_code": 80141501, "path": "Segment > ... > Commodity", "reason": "why it matches"}
  ],
  "analysis_rows": [
     {"icon": "inventory_2", "left": "...", "right": "FOUND", "right_style": "success"},
     {"icon": "verified_user", "left": "...", "right": "CDW, B&H", "right_style": "muted"}
  ],
  "form_intro": "One paragraph telling the user to confirm line items for the selected catalogue item."
}

Use right_style: "success" (green), "muted" (grey), or "neutral".
Icons must be Material symbol names: inventory_2, verified_user, travel_explore, psychology, etc.
"""


TOOLS: list[dict[str, Any]] = [
    {
        "type": "function",
        "function": {
            "name": "search_catalog",
            "description": "Keyword search in the SQLite procurement catalogue (FTS). Pass 2–6 focused procurement keywords (e.g. market research, survey services). Call again with broader alternate phrases if the first query returns few or zero rows.",
            "parameters": {
                "type": "object",
                "properties": {
                    "query": {
                        "type": "string",
                        "description": "Focused search phrase, e.g. 'market research laptop depot repair'",
                    },
                    "limit": {
                        "type": "integer",
                        "description": "Max rows to return (default 20)",
                    },
                },
                "required": ["query"],
            },
        },
    },
    {
        "type": "function",
        "function": {
            "name": "get_commodity_details",
            "description": "Fetch one commodity by exact 8-digit catalogue commodity code.",
            "parameters": {
                "type": "object",
                "properties": {
                    "commodity_code": {
                        "type": "integer",
                        "description": "Catalogue commodity code, e.g. 80141501",
                    }
                },
                "required": ["commodity_code"],
            },
        },
    },
]


def _dispatch(
    conn: sqlite3.Connection, name: str, args: dict[str, Any]
) -> Any:
    if name == "search_catalog":
        q = str(args.get("query") or "")
        limit = int(args.get("limit") or 20)
        rows = search_catalog(conn, q, limit=limit)
        return {"matches": [summarize_row(r) for r in rows]}
    if name == "get_commodity_details":
        code = int(args.get("commodity_code"))
        row = get_commodity(conn, code)
        if not row:
            return {"error": "not_found", "commodity_code": code}
        return {"commodity": summarize_row(row)}
    return {"error": "unknown_tool", "name": name}


def run_agent(
    user_message: str,
    *,
    conn: sqlite3.Connection | None = None,
    selected_code: int | None = None,
) -> dict[str, Any]:
    """Return parsed JSON object from model after tool loop."""
    own_conn = conn is None
    if conn is None:
        conn = connect()

    try:
        if selected_code is not None:
            row = get_commodity(conn, int(selected_code))
            if not row:
                return {
                    "status": "not_found",
                    "summary": "That commodity code is not in the local catalog.",
                    "commodity_code": None,
                    "candidates": [],
                    "analysis_rows": [],
                    "form_intro": "",
                }
            s = summarize_row(row)
            return {
                "status": "found",
                "summary": f"Selected catalogue item {s['commodity_code']}: {s['path']}",
                "commodity_code": s["commodity_code"],
                "candidates": [],
                "analysis_rows": [
                    {
                        "icon": "inventory_2",
                        "left": f"Catalog match: {s['path']}",
                        "right": "FOUND",
                        "right_style": "success",
                    },
                    {
                        "icon": "verified_user",
                        "left": "Cross-referencing approved vendors (configure in procurement policy)",
                        "right": "—",
                        "right_style": "muted",
                    },
                ],
                "form_intro": (
                    "Review the specification fields for this catalogue item. "
                    "Confirm selections before generating the requisition."
                ),
                "selected_details": s,
            }

        api_key = os.environ.get("OPENAI_API_KEY")
        if not api_key:
            return {
                "status": "not_found",
                "summary": "Server misconfiguration: OPENAI_API_KEY is not set.",
                "commodity_code": None,
                "candidates": [],
                "analysis_rows": [],
                "form_intro": "",
                "error": "missing_openai_key",
            }

        client = OpenAI(api_key=api_key)
        messages: list[dict[str, Any]] = [
            {"role": "system", "content": SYSTEM},
            {"role": "user", "content": user_message},
        ]

        for _ in range(10):
            resp = client.chat.completions.create(
                model=os.environ.get("OPENAI_MODEL", "gpt-4o-mini"),
                messages=messages,
                tools=TOOLS,
                tool_choice="auto",
                temperature=0.2,
            )
            msg = resp.choices[0].message
            if msg.tool_calls:
                messages.append(_assistant_message_dict(msg))
                for tc in msg.tool_calls:
                    name = tc.function.name
                    try:
                        args = json.loads(tc.function.arguments or "{}")
                    except json.JSONDecodeError:
                        args = {}
                    result = _dispatch(conn, name, args)
                    messages.append(
                        {
                            "role": "tool",
                            "tool_call_id": tc.id,
                            "content": json.dumps(result, ensure_ascii=False),
                        }
                    )
                continue

            text = (msg.content or "").strip()
            parsed = _parse_json_response(text)
            if parsed:
                return parsed
            return {
                "status": "not_found",
                "summary": text or "Unable to parse agent response.",
                "commodity_code": None,
                "candidates": [],
                "analysis_rows": [],
                "form_intro": "",
                "raw": text,
            }

        return {
            "status": "not_found",
            "summary": "Agent stopped after too many tool rounds.",
            "commodity_code": None,
            "candidates": [],
            "analysis_rows": [],
            "form_intro": "",
        }
    finally:
        if own_conn:
            conn.close()


def coerce_payload(payload: dict[str, Any]) -> dict[str, Any]:
    """Ensure required keys exist."""
    base = {
        "status": payload.get("status") or "not_found",
        "summary": payload.get("summary") or "",
        "commodity_code": payload.get("commodity_code"),
        "candidates": payload.get("candidates") or [],
        "analysis_rows": payload.get("analysis_rows") or [],
        "form_intro": payload.get("form_intro") or "",
    }
    if "selected_details" in payload:
        base["selected_details"] = payload["selected_details"]
    if "error" in payload:
        base["error"] = payload["error"]
    return base