File size: 11,711 Bytes
e09a32e
 
 
 
 
 
 
 
 
28b1c4e
e09a32e
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
277c5a2
 
 
 
 
 
 
 
 
 
 
 
e09a32e
7a36038
277c5a2
7a36038
 
 
e09a32e
 
 
 
7a36038
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
e09a32e
7a36038
 
e09a32e
7a36038
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
e09a32e
7a36038
 
e09a32e
 
7a36038
e09a32e
7a36038
 
 
 
 
 
 
 
 
e09a32e
7a36038
 
 
e09a32e
7a36038
e09a32e
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
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
# app.py — veureu/stools (Salamandra 7B Tools · ZeroGPU) — compatible con ENGINE
from __future__ import annotations
import os, json, re
from typing import List, Dict, Any, Optional, Tuple

import gradio as gr
import spaces
import torch
from transformers import AutoTokenizer, AutoModelForCausalLM
from salamandra_tools import SalamandraClient

# ================= Config =================
MODEL_ID = os.environ.get("MODEL_ID", "BSC-LT/salamandra-7b-tools")
DTYPE = torch.float16 if torch.cuda.is_available() else torch.float32
DEVICE = "cuda" if torch.cuda.is_available() else "cpu"

_tok = None
_model = None

def _lazy_load() -> Tuple[AutoTokenizer, AutoModelForCausalLM]:
    global _tok, _model
    if _tok is None or _model is None:
        _tok = AutoTokenizer.from_pretrained(MODEL_ID, use_fast=True, trust_remote_code=True)
        _model = AutoModelForCausalLM.from_pretrained(
            MODEL_ID,
            torch_dtype=DTYPE,
            low_cpu_mem_usage=True,
            use_safetensors=True,
            trust_remote_code=True,
            device_map=None,
        ).to(DEVICE)
    return _tok, _model


# =============== Helpers ===============

def _render_tools_md(tools: List[Dict[str, Any]]) -> str:
    """Convierte la especificación OpenAI-style de tools a un bloque breve markdown para el prompt."""
    if not tools:
        return ""
    lines = ["Herramientas disponibles (formato JSON):"]
    for t in tools:
        name = t.get("function", {}).get("name") or t.get("name") or "tool"
        desc = t.get("function", {}).get("description") or t.get("description") or ""
        params = t.get("function", {}).get("parameters") or t.get("parameters") or {}
        lines.append(f"- **{name}**: {desc} | parámetros: {json.dumps(params)[:600]}")
    return "\n".join(lines)

def _compose_chat_prompt(messages: List[Dict[str, str]], tools_md: str) -> str:
    """
    Soporta mensajes estilo OpenAI: [{"role":"system|user|assistant", "content":"..."}]
    Usa chat_template si está disponible.
    """
    tok, _ = _lazy_load()
    sys_text = ""
    usr_msgs: List[Dict[str, str]] = []
    for m in messages:
        role = m.get("role", "")
        content = (m.get("content") or "").strip()
        if role == "system":
            sys_text += ("\n" + content) if sys_text else content
        else:
            usr_msgs.append({"role": role, "content": content})

    # injerta descripción de tools en el system
    if tools_md:
        sys_text = (sys_text + "\n\n" if sys_text else "") + tools_md + \
                   "\n\nSi decides llamar a una herramienta, devuelve un objeto JSON con la clave 'tool_calls' " \
                   "y describe tus razonamientos de forma concisa en 'thought' (opcional)."

    # reconstruimos la conversación con system delante
    conv: List[Dict[str, str]] = []
    if sys_text:
        conv.append({"role":"system", "content": sys_text})
    conv.extend(usr_msgs)

    chat_template = getattr(tok, "chat_template", None)
    if chat_template:
        return tok.apply_chat_template(conv, tokenize=False, add_generation_prompt=True)

    # Fallback sin plantilla
    rendered = ""
    if sys_text:
        rendered += f"<<SYS>>\n{sys_text}\n<</SYS>>\n\n"
    for m in usr_msgs:
        if m["role"] == "user":
            rendered += f"### Usuario\n{m['content']}\n\n"
        elif m["role"] == "assistant":
            rendered += f"### Asistente\n{m['content']}\n\n"
    rendered += "### Asistente\n"
    return rendered


# =============== (Opcional) Mini-ejecutor local de herramientas seguras ===============
# Si el LLM devuelve {"tool_calls":[{"name":"calculator","arguments":{"expr":"2+2"}}]}
# podemos ejecutar algunas herramientas inofensivas de ejemplo.
# Nota: mantén esto muy simple/seguro. Puedes desactivarlo poniendo EXECUTE_TOOLS=False.
EXECUTE_TOOLS = True

def _safe_calculator(expr: str) -> str:
    # Permite solo dígitos, espacios, (), y +-*/.%**
    if not re.fullmatch(r"[0-9\.\s\+\-\*\/\%\(\)\^eE]+", expr.replace("**","^")):
        return "Rejected expression."
    # soporta ^ como potencia -> **
    expr = expr.replace("^", "**")
    try:
        return str(eval(expr, {"__builtins__":{}}, {}))
    except Exception as e:
        return f"Error: {e}"

LOCAL_TOOLBOX = {
    "calculator": lambda args: _safe_calculator(str(args.get("expr",""))),
}

def maybe_execute_tool_calls(tool_calls: List[Dict[str, Any]]) -> List[Dict[str, Any]]:
    if not EXECUTE_TOOLS:
        return []
    results = []
    for call in tool_calls:
        name = call.get("name")
        args = call.get("arguments", {})
        fn = LOCAL_TOOLBOX.get(name)
        if fn is None:
            results.append({"name": name, "error": "tool_not_available"})
            continue
        try:
            out = fn(args)
            results.append({"name": name, "output": out})
        except Exception as e:
            results.append({"name": name, "error": str(e)})
    return results


# =============== Core generation ===============

@spaces.GPU  # usa GPU si está disponible (ZeroGPU)
def _generate_with_tools(
    messages: List[Dict[str, str]],
    tools: List[Dict[str, Any]],
    max_new_tokens: int = 512,
    temperature: float = 0.7,
    top_p: float = 0.95,
) -> Dict[str, Any]:
    tok, model = _lazy_load()
    tools_md = _render_tools_md(tools)
    prompt = _compose_chat_prompt(messages, tools_md)

    inputs = tok(prompt, return_tensors="pt").to(DEVICE)
    with torch.inference_mode():
        out = model.generate(
            **inputs,
            max_new_tokens=int(max_new_tokens),
            temperature=float(temperature),
            top_p=float(top_p),
            do_sample=True if temperature > 0 else False,
            pad_token_id=tok.eos_token_id,
            eos_token_id=tok.eos_token_id,
        )
    text = tok.decode(out[0], skip_special_tokens=True).strip()

    # Si el modelo devuelve un bloque JSON con 'tool_calls', lo intentamos extraer.
    tool_calls: List[Dict[str, Any]] = []
    try:
        # busca el último {...} que contenga "tool_calls"
        matches = list(re.finditer(r"\{.*?\"tool_calls\".*?\}", text, flags=re.S))
        if matches:
            block = text[matches[-1].start():matches[-1].end()]
            obj = json.loads(block)
            tc = obj.get("tool_calls", [])
            if isinstance(tc, list):
                tool_calls = tc
    except Exception:
        pass

    tool_results = maybe_execute_tool_calls(tool_calls) if tool_calls else []

    return {"text": text, "tool_calls": tool_calls, "tool_results": tool_results}


# =================== Gradio Endpoints ===================

def predict_for_engine(messages_json: str, tools_json: str) -> Dict[str, Any]:
    """
    Endpoint esperado por ENGINE (ToolsClient.chat):
      - messages_json: JSON de [{"role":"user|assistant|system","content":"..."}]
      - tools_json: JSON OpenAI-like de herramientas (opcional)
    Devuelve: {"text": "...", "tool_calls": [...], "tool_results": [...]}
    """
    try:
        messages = json.loads(messages_json) if messages_json else []
    except Exception:
        messages = []
    try:
        tools = json.loads(tools_json) if tools_json else []
    except Exception:
        tools = []
    return _generate_with_tools(messages, tools, max_new_tokens=512, temperature=0.7, top_p=0.95)

def chat_advanced(messages_json: str, tools_json: str, max_new_tokens: int, temperature: float, top_p: float) -> Dict[str, Any]:
    try:
        messages = json.loads(messages_json) if messages_json else []
    except Exception:
        messages = []
    try:
        tools = json.loads(tools_json) if tools_json else []
    except Exception:
        tools = []
    return _generate_with_tools(messages, tools, max_new_tokens=int(max_new_tokens), temperature=float(temperature), top_p=float(top_p))


_salamandra = None

def salamandra_chat_endpoint(prompt: str) -> Dict[str, Any]:
    global _salamandra
    if _salamandra is None:
        _salamandra = SalamandraClient()   # usa tu clase

    try:
        text = _salamandra.chat(prompt)
    except Exception as e:
        text = f"Error ejecutando SalamandraClient: {str(e)}"

    return {"text": text}
    
# =================== UI ===================
custom_css = """
h2 {
    background: #e3e4e6 !important;
    padding: 14px 22px !important;
    border-radius: 14px !important;
    box-shadow: 0 4px 12px rgba(0,0,0,0.08) !important;
    display: block !important;       /* ocupa tot l'ample */
    width: 100% !important;          /* assegura 100% */
    margin: 20px auto !important;
    text-align:center;
}
"""

# Main interface for Salamandra 7B Tools: supports tool specification (function calling)
with gr.Blocks(title="Salamandra 7B Tools · ZeroGPU", css=custom_css, theme=gr.themes.Soft()) as demo:

    # Header description for the UI
    gr.Markdown("## Salamandra-7B-Tools · ZeroGPU\nXat amb especificació d'eines (function-calling).")

    with gr.Row():
        with gr.Column():

            # JSON array of messages passed to the model
            messages = gr.Textbox(
                label="Missatges (JSON)",
                value='[{"role":"user","content":"Quant és (2+2)^3?"}]',
                lines=6
            )

            # Tool definitions in JSON schema format
            tools = gr.Textbox(
                label="Eines (JSON, opcional)",
                value='[{"type":"function","function":{"name":"calculator","description":"Avalua expressions aritmètiques bàsiques.","parameters":{"type":"object","properties":{"expr":{"type":"string"}},"required":["expr"]}}}]',
                lines=6
            )

            # Maximum generation length
            max_new = gr.Slider(16, 2048, value=512, step=16, label="Màxim de tokens nous")

            # Temperature for randomness
            temp = gr.Slider(0.0, 1.5, value=0.7, step=0.05, label="Temperatura")

            # Nucleus sampling threshold
            topp = gr.Slider(0.1, 1.0, value=0.95, step=0.01, label="Top-p")

            # Button to generate a response
            btn = gr.Button("Generar", variant="primary")

        with gr.Column():
            # JSON output from the model
            out = gr.JSON(label="Sortida")

    # Bind chat-with-tools generation
    btn.click(
        chat_advanced,
        [messages, tools, max_new, temp, topp],
        out,
        api_name="chat",
        concurrency_limit=1
    )

    # --------------------------------------------------------------
    gr.Markdown("---")
    # --------------------------------------------------------------

    # Minimal /predict endpoint for ENGINE compatibility
    # Accepts messages + tool definitions
    gr.Button("Provar /predict").click(
        predict_for_engine,
        [messages, tools],
        out,
        api_name="predict",
        concurrency_limit=1
    )

    # --------------------------------------------------------------
    gr.Markdown("---")
    # --------------------------------------------------------------

    # Endpoint: raw prompt → model output (JSON)
    with gr.Row():
        prompt = gr.Textbox(label="Prompt", lines=10)

    with gr.Row():
        btn2 = gr.Button("Generar", variant="primary")

    with gr.Row():
        out2 = gr.JSON(label="Sortida")

    btn2.click(
        salamandra_chat_endpoint,
        [prompt],
        out2,
        api_name="generate_out_from_prompt",
        concurrency_limit=1
    )

    # --------------------------------------------------------------
    gr.Markdown("---")
    # --------------------------------------------------------------

# Enable request queue for concurrency safety
demo.queue(max_size=16).launch()