superlocalmemory 3.0.17 → 3.0.18

This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
@@ -12,7 +12,7 @@ from typing import Optional
12
12
  from fastapi import APIRouter, HTTPException, Query, Request
13
13
 
14
14
  from .helpers import (
15
- get_db_connection, dict_factory, get_active_profile,
15
+ get_db_connection, dict_factory, get_active_profile, get_engine_lazy,
16
16
  SearchRequest, DB_PATH, MEMORY_DIR,
17
17
  )
18
18
 
@@ -21,8 +21,8 @@ router = APIRouter()
21
21
 
22
22
 
23
23
  def _get_engine(request: Request):
24
- """Get V3 engine from app state, or None."""
25
- return getattr(request.app.state, "engine", None)
24
+ """Get V3 engine from app state, initializing lazily on first call."""
25
+ return get_engine_lazy(request.app.state)
26
26
 
27
27
 
28
28
  def _preview(content: str | None) -> str:
@@ -202,7 +202,7 @@ async def get_memories(
202
202
 
203
203
  if use_v3:
204
204
  query = """
205
- SELECT fact_id as id, content, fact_type as category,
205
+ SELECT fact_id as id, memory_id, content, fact_type as category,
206
206
  confidence as importance, access_count,
207
207
  created_at, created_at as updated_at,
208
208
  session_id as project_name
@@ -313,39 +313,22 @@ async def get_graph(
313
313
 
314
314
  @router.post("/api/search")
315
315
  async def search_memories(request: Request, body: SearchRequest):
316
- """Semantic search using V3 engine recall or fallback."""
316
+ """Semantic search via subprocess worker pool (memory-isolated)."""
317
317
  try:
318
- engine = _get_engine(request)
319
-
320
- if engine:
321
- response = engine.recall(body.query, limit=body.limit)
322
- results = []
323
- for r in response.results:
324
- score = r.score
325
- if score < body.min_score:
326
- continue
327
- if body.category and getattr(r.fact, 'fact_type', None) != body.category:
328
- continue
329
- results.append({
330
- "id": r.fact.fact_id,
331
- "content": r.fact.content,
332
- "score": round(score, 4),
333
- "confidence": round(r.confidence, 4),
334
- "trust_score": round(r.trust_score, 4) if r.trust_score else None,
335
- "channel_scores": r.channel_scores,
336
- "fact_type": getattr(r.fact, 'fact_type', None),
337
- "created_at": getattr(r.fact, 'created_at', None),
338
- })
339
- if len(results) >= body.limit:
340
- break
318
+ from superlocalmemory.core.worker_pool import WorkerPool
319
+ pool = WorkerPool.shared()
320
+ result = pool.recall(body.query, limit=body.limit)
341
321
 
322
+ if result.get("ok"):
342
323
  return {
343
- "query": body.query, "results": results, "total": len(results),
344
- "query_type": response.query_type,
345
- "retrieval_time_ms": response.retrieval_time_ms,
324
+ "query": body.query,
325
+ "results": result.get("results", []),
326
+ "total": result.get("result_count", 0),
327
+ "query_type": result.get("query_type", "unknown"),
328
+ "retrieval_time_ms": result.get("retrieval_time_ms", 0),
346
329
  }
347
330
 
348
- # Fallback: direct DB search (no V3 engine)
331
+ # Fallback: direct DB text search (no engine needed)
349
332
  conn = get_db_connection()
350
333
  conn.row_factory = dict_factory
351
334
  cursor = conn.cursor()
@@ -434,8 +417,41 @@ async def get_cluster_detail(request: Request, cluster_id: int, limit: int = Que
434
417
  conn.close()
435
418
  if not members:
436
419
  raise HTTPException(status_code=404, detail="Cluster not found")
437
- return {"cluster_info": {"cluster_id": cluster_id, "total_members": len(members)}, "members": members, "connections": []}
420
+ # Generate cluster summary
421
+ summary = ""
422
+ try:
423
+ from superlocalmemory.core.worker_pool import WorkerPool
424
+ pool = WorkerPool.shared()
425
+ texts = [m.get("content", "")[:200] for m in members[:10] if m.get("content")]
426
+ if texts:
427
+ result = pool.summarize(texts)
428
+ summary = result.get("summary", "") if result.get("ok") else ""
429
+ except Exception:
430
+ pass
431
+
432
+ return {
433
+ "cluster_info": {"cluster_id": cluster_id, "total_members": len(members)},
434
+ "summary": summary,
435
+ "members": members,
436
+ "connections": [],
437
+ }
438
438
  except HTTPException:
439
439
  raise
440
440
  except Exception as e:
441
441
  raise HTTPException(status_code=500, detail=f"Cluster detail error: {str(e)}")
442
+
443
+
444
+ @router.get("/api/memories/{memory_id}/facts")
445
+ async def get_memory_facts(request: Request, memory_id: str):
446
+ """Get original memory text with all its child atomic facts."""
447
+ try:
448
+ from superlocalmemory.core.worker_pool import WorkerPool
449
+ pool = WorkerPool.shared()
450
+ result = pool.get_memory_facts(memory_id)
451
+ if result.get("ok"):
452
+ return result
453
+ raise HTTPException(status_code=404, detail="Memory not found")
454
+ except HTTPException:
455
+ raise
456
+ except Exception as e:
457
+ raise HTTPException(status_code=500, detail=f"Error: {str(e)}")
@@ -8,6 +8,7 @@ from __future__ import annotations
8
8
 
9
9
  import json
10
10
  import logging
11
+ import os
11
12
  from pathlib import Path
12
13
  from fastapi import APIRouter, Request
13
14
  from fastapi.responses import JSONResponse
@@ -63,17 +64,22 @@ async def dashboard(request: Request):
63
64
 
64
65
  @router.get("/mode")
65
66
  async def get_mode():
66
- """Get current operating mode."""
67
+ """Get current mode, provider, model — single source of truth for UI."""
67
68
  try:
68
69
  from superlocalmemory.core.config import SLMConfig
69
70
  config = SLMConfig.load()
70
- modes = {
71
- "a": {"name": "Local Guardian", "description": "Zero cloud. Your data never leaves your machine.", "llm": False, "eu_compliant": True},
72
- "b": {"name": "Smart Local", "description": "Local LLM via Ollama. Still fully private.", "llm": "local", "eu_compliant": True},
73
- "c": {"name": "Full Power", "description": "Cloud LLM for maximum accuracy.", "llm": "cloud", "eu_compliant": False},
74
- }
75
71
  current = config.mode.value
76
- return {"current": current, "details": modes.get(current, {}), "all_modes": modes}
72
+ return {
73
+ "mode": current,
74
+ "provider": config.llm.provider or "none",
75
+ "model": config.llm.model or "",
76
+ "has_key": bool(config.llm.api_key),
77
+ "endpoint": config.llm.api_base or "",
78
+ "capabilities": {
79
+ "llm_available": bool(config.llm.provider),
80
+ "cross_encoder": config.retrieval.use_cross_encoder if hasattr(config, 'retrieval') else False,
81
+ },
82
+ }
77
83
  except Exception as e:
78
84
  return JSONResponse({"error": str(e)}, status_code=500)
79
85
 
@@ -109,6 +115,127 @@ async def set_mode(request: Request):
109
115
  return JSONResponse({"error": str(e)}, status_code=500)
110
116
 
111
117
 
118
+ @router.post("/mode/set")
119
+ async def set_full_config(request: Request):
120
+ """Save mode + provider + model + API key together."""
121
+ try:
122
+ body = await request.json()
123
+ new_mode = body.get("mode", "a").lower()
124
+ provider = body.get("provider", "none")
125
+ model = body.get("model", "")
126
+ api_key = body.get("api_key", "")
127
+
128
+ if new_mode not in ("a", "b", "c"):
129
+ return JSONResponse({"error": "Invalid mode"}, status_code=400)
130
+
131
+ from superlocalmemory.core.config import SLMConfig
132
+ from superlocalmemory.storage.models import Mode
133
+ config = SLMConfig.for_mode(
134
+ Mode(new_mode),
135
+ llm_provider=provider if provider != "none" else "",
136
+ llm_model=model,
137
+ llm_api_key=api_key,
138
+ llm_api_base="http://localhost:11434" if provider == "ollama" else "",
139
+ )
140
+ old = SLMConfig.load()
141
+ config.active_profile = old.active_profile
142
+ config.save()
143
+
144
+ # Kill existing worker so next request uses new config
145
+ try:
146
+ from superlocalmemory.core.worker_pool import WorkerPool
147
+ WorkerPool.shared().shutdown()
148
+ except Exception:
149
+ pass
150
+
151
+ if hasattr(request.app.state, "engine"):
152
+ request.app.state.engine = None
153
+
154
+ return {
155
+ "success": True,
156
+ "mode": new_mode,
157
+ "provider": provider,
158
+ "model": model,
159
+ }
160
+ except Exception as e:
161
+ return JSONResponse({"error": str(e)}, status_code=500)
162
+
163
+
164
+ @router.post("/provider/test")
165
+ async def test_provider(request: Request):
166
+ """Test connectivity to an LLM provider."""
167
+ try:
168
+ import httpx
169
+ body = await request.json()
170
+ provider = body.get("provider", "")
171
+ model = body.get("model", "")
172
+ api_key = body.get("api_key", "")
173
+
174
+ if provider == "ollama":
175
+ endpoint = body.get("endpoint", "http://localhost:11434")
176
+ with httpx.Client(timeout=httpx.Timeout(5.0)) as c:
177
+ resp = c.get(f"{endpoint}/api/tags")
178
+ resp.raise_for_status()
179
+ models = [m["name"] for m in resp.json().get("models", [])]
180
+ found = model in models if model else len(models) > 0
181
+ return {
182
+ "success": found,
183
+ "message": f"Ollama OK, {len(models)} models" + (f", '{model}' available" if found and model else ""),
184
+ }
185
+
186
+ if provider == "openrouter":
187
+ if not api_key:
188
+ api_key = os.environ.get("OPENROUTER_API_KEY", "")
189
+ if not api_key:
190
+ return {"success": False, "error": "API key required"}
191
+ with httpx.Client(timeout=httpx.Timeout(10.0)) as c:
192
+ resp = c.get("https://openrouter.ai/api/v1/models", headers={"Authorization": f"Bearer {api_key}"})
193
+ resp.raise_for_status()
194
+ return {"success": True, "message": "OpenRouter connected, key valid"}
195
+
196
+ if provider == "openai":
197
+ if not api_key:
198
+ return {"success": False, "error": "API key required"}
199
+ with httpx.Client(timeout=httpx.Timeout(10.0)) as c:
200
+ resp = c.get("https://api.openai.com/v1/models", headers={"Authorization": f"Bearer {api_key}"})
201
+ resp.raise_for_status()
202
+ return {"success": True, "message": "OpenAI connected, key valid"}
203
+
204
+ if provider == "anthropic":
205
+ if not api_key:
206
+ return {"success": False, "error": "API key required"}
207
+ # Anthropic doesn't have a models list endpoint, just verify key format
208
+ if api_key.startswith("sk-ant-"):
209
+ return {"success": True, "message": "Anthropic key format valid"}
210
+ return {"success": False, "error": "Key should start with sk-ant-"}
211
+
212
+ return {"success": False, "error": f"Unknown provider: {provider}"}
213
+ except httpx.ConnectError:
214
+ return {"success": False, "error": "Cannot connect — is the service running?"}
215
+ except httpx.HTTPStatusError as e:
216
+ return {"success": False, "error": f"HTTP {e.response.status_code}: Invalid key or endpoint"}
217
+ except Exception as e:
218
+ return {"success": False, "error": str(e)}
219
+
220
+
221
+ @router.get("/ollama/status")
222
+ async def ollama_status():
223
+ """Check if Ollama is running and list available models."""
224
+ try:
225
+ import httpx
226
+ with httpx.Client(timeout=httpx.Timeout(5.0)) as client:
227
+ resp = client.get("http://localhost:11434/api/tags")
228
+ resp.raise_for_status()
229
+ data = resp.json()
230
+ models = [
231
+ {"name": m["name"], "size": m.get("size", 0)}
232
+ for m in data.get("models", [])
233
+ ]
234
+ return {"running": True, "models": models, "count": len(models)}
235
+ except Exception:
236
+ return {"running": False, "models": [], "count": 0}
237
+
238
+
112
239
  # ── Provider ─────────────────────────────────────────────────
113
240
 
114
241
  @router.get("/providers")
@@ -187,28 +314,32 @@ async def recall_trace(request: Request):
187
314
  query = body.get("query", "")
188
315
  limit = body.get("limit", 10)
189
316
 
190
- engine = getattr(request.app.state, "engine", None)
191
- if not engine:
192
- return JSONResponse({"error": "Engine not initialized"}, status_code=503)
193
-
194
- response = engine.recall(query, limit=limit)
195
- results = []
196
- for r in response.results[:limit]:
197
- results.append({
198
- "fact_id": r.fact.fact_id,
199
- "content": r.fact.content[:300],
200
- "score": round(r.score, 4),
201
- "confidence": round(r.confidence, 4),
202
- "trust_score": round(r.trust_score, 4),
203
- "channel_scores": {k: round(v, 4) for k, v in (r.channel_scores or {}).items()},
204
- })
317
+ from superlocalmemory.core.worker_pool import WorkerPool
318
+ pool = WorkerPool.shared()
319
+ result = pool.recall(query, limit=limit)
320
+
321
+ if not result.get("ok"):
322
+ return JSONResponse(
323
+ {"error": result.get("error", "Recall failed")},
324
+ status_code=503,
325
+ )
326
+
327
+ # Optional: synthesize answer from results (Mode B/C only)
328
+ synthesis = ""
329
+ if body.get("synthesize") and result.get("results"):
330
+ try:
331
+ syn_result = pool.synthesize(query, result["results"][:5])
332
+ synthesis = syn_result.get("synthesis", "") if syn_result.get("ok") else ""
333
+ except Exception:
334
+ pass
205
335
 
206
336
  return {
207
337
  "query": query,
208
- "query_type": response.query_type,
209
- "result_count": len(results),
210
- "retrieval_time_ms": round(response.retrieval_time_ms, 1),
211
- "results": results,
338
+ "query_type": result.get("query_type", "unknown"),
339
+ "result_count": result.get("result_count", 0),
340
+ "retrieval_time_ms": result.get("retrieval_time_ms", 0),
341
+ "results": result.get("results", []),
342
+ "synthesis": synthesis,
212
343
  }
213
344
  except Exception as e:
214
345
  return JSONResponse({"error": str(e)}, status_code=500)
@@ -218,29 +349,50 @@ async def recall_trace(request: Request):
218
349
 
219
350
  @router.get("/trust/dashboard")
220
351
  async def trust_dashboard(request: Request):
221
- """Trust overview: per-agent scores, alerts."""
352
+ """Trust overview: per-agent scores, alerts. Queries DB directly."""
222
353
  try:
223
- engine = getattr(request.app.state, "engine", None)
224
- if not engine or not engine._trust_scorer:
225
- return {"agents": [], "alerts": [], "message": "Trust scorer not available"}
226
-
227
354
  from superlocalmemory.core.config import SLMConfig
355
+ from superlocalmemory.storage.database import DatabaseManager
356
+ from superlocalmemory.storage import schema as _schema
228
357
  config = SLMConfig.load()
229
- scores = engine._trust_scorer.get_all_scores(config.active_profile)
358
+ pid = config.active_profile
359
+
360
+ db_path = config.db_path
361
+ db = DatabaseManager(db_path)
362
+ db.initialize(_schema)
230
363
 
364
+ # Query trust scores from DB
231
365
  agents = []
232
- for s in scores:
233
- if isinstance(s, dict):
234
- agents.append(s)
235
- else:
366
+ try:
367
+ rows = db.execute(
368
+ "SELECT target_id, target_type, trust_score, evidence_count, "
369
+ "last_updated FROM trust_scores WHERE profile_id = ? "
370
+ "ORDER BY trust_score DESC",
371
+ (pid,),
372
+ )
373
+ for r in rows:
374
+ d = dict(r)
236
375
  agents.append({
237
- "target_id": s.target_id,
238
- "target_type": s.target_type,
239
- "trust_score": round(s.trust_score, 3),
240
- "evidence_count": s.evidence_count,
376
+ "target_id": d.get("target_id", ""),
377
+ "target_type": d.get("target_type", ""),
378
+ "trust_score": round(float(d.get("trust_score", 0.5)), 3),
379
+ "evidence_count": d.get("evidence_count", 0),
380
+ "last_updated": d.get("last_updated", ""),
241
381
  })
382
+ except Exception:
383
+ pass
384
+
385
+ # Aggregate stats
386
+ avg = round(sum(a["trust_score"] for a in agents) / len(agents), 3) if agents else 0.5
387
+ alerts = [a for a in agents if a["trust_score"] < 0.3]
242
388
 
243
- return {"agents": agents, "alerts": [], "profile": config.active_profile}
389
+ return {
390
+ "agents": agents,
391
+ "avg_trust": avg,
392
+ "alerts": alerts,
393
+ "total": len(agents),
394
+ "profile": pid,
395
+ }
244
396
  except Exception as e:
245
397
  return JSONResponse({"error": str(e)}, status_code=500)
246
398
 
@@ -249,9 +401,9 @@ async def trust_dashboard(request: Request):
249
401
 
250
402
  @router.get("/math/health")
251
403
  async def math_health(request: Request):
252
- """Mathematical layer health: Fisher, sheaf, Langevin status."""
404
+ """Mathematical layer health: Fisher, sheaf, Langevin status. Queries DB directly."""
253
405
  try:
254
- engine = getattr(request.app.state, "engine", None)
406
+ engine = None # Engine runs in subprocess; query DB directly below
255
407
 
256
408
  health = {
257
409
  "fisher": {"status": "active", "description": "Fisher-Rao information geometry for similarity"},
@@ -196,22 +196,23 @@ def create_app() -> FastAPI:
196
196
 
197
197
  @application.on_event("startup")
198
198
  async def startup_event():
199
- """Initialize V3 engine and event bus on startup."""
200
- # Initialize V3 engine for dashboard API routes
201
- try:
202
- from superlocalmemory.core.config import SLMConfig
203
- from superlocalmemory.core.engine import MemoryEngine
204
- config = SLMConfig.load()
205
- engine = MemoryEngine(config)
206
- engine.initialize()
207
- application.state.engine = engine
208
- logger.info("V3 engine initialized for dashboard")
209
- except Exception as exc:
210
- logger.warning("V3 engine init failed: %s (V3 API routes will be unavailable)", exc)
211
- application.state.engine = None
212
-
199
+ """Initialize event bus. Engine runs in subprocess worker (never in this process)."""
200
+ # Engine is NEVER loaded in the dashboard process.
201
+ # All recall/search operations go through WorkerPool subprocess.
202
+ # This keeps the dashboard permanently at ~60 MB.
203
+ application.state.engine = None
204
+ logger.info("Dashboard started (~60 MB, engine runs in subprocess worker)")
213
205
  register_event_listener()
214
206
 
207
+ @application.on_event("shutdown")
208
+ async def shutdown_event():
209
+ """Kill worker subprocess on dashboard shutdown."""
210
+ try:
211
+ from superlocalmemory.core.worker_pool import WorkerPool
212
+ WorkerPool.shared().shutdown()
213
+ except Exception:
214
+ pass
215
+
215
216
  return application
216
217
 
217
218
 
@@ -300,6 +300,29 @@ class DatabaseManager:
300
300
  for r in rows
301
301
  ]
302
302
 
303
+ def get_memory_content_batch(self, memory_ids: list[str]) -> dict[str, str]:
304
+ """Batch-fetch original memory text. Returns {memory_id: content}."""
305
+ if not memory_ids:
306
+ return {}
307
+ unique_ids = list(set(memory_ids))
308
+ ph = ','.join('?' * len(unique_ids))
309
+ rows = self.execute(
310
+ f"SELECT memory_id, content FROM memories WHERE memory_id IN ({ph})",
311
+ tuple(unique_ids),
312
+ )
313
+ return {dict(r)["memory_id"]: dict(r)["content"] for r in rows}
314
+
315
+ def get_facts_by_memory_id(
316
+ self, memory_id: str, profile_id: str,
317
+ ) -> list[AtomicFact]:
318
+ """Get all atomic facts for a given memory_id."""
319
+ rows = self.execute(
320
+ "SELECT * FROM atomic_facts WHERE memory_id = ? AND profile_id = ? "
321
+ "ORDER BY confidence DESC",
322
+ (memory_id, profile_id),
323
+ )
324
+ return [self._row_to_fact(r) for r in rows]
325
+
303
326
  def store_edge(self, edge: GraphEdge) -> str:
304
327
  """Persist a graph edge. Returns edge_id."""
305
328
  self.execute(
@@ -1,6 +1,6 @@
1
1
  Metadata-Version: 2.4
2
2
  Name: superlocalmemory
3
- Version: 3.0.17
3
+ Version: 3.0.18
4
4
  Summary: Information-geometric agent memory with mathematical guarantees
5
5
  Author-email: Varun Pratap Bhardwaj <admin@superlocalmemory.com>
6
6
  License: MIT
@@ -30,13 +30,17 @@ src/superlocalmemory/compliance/retention.py
30
30
  src/superlocalmemory/compliance/scheduler.py
31
31
  src/superlocalmemory/core/__init__.py
32
32
  src/superlocalmemory/core/config.py
33
+ src/superlocalmemory/core/embedding_worker.py
33
34
  src/superlocalmemory/core/embeddings.py
34
35
  src/superlocalmemory/core/engine.py
35
36
  src/superlocalmemory/core/hooks.py
36
37
  src/superlocalmemory/core/maintenance.py
37
38
  src/superlocalmemory/core/modes.py
38
39
  src/superlocalmemory/core/profiles.py
40
+ src/superlocalmemory/core/recall_worker.py
39
41
  src/superlocalmemory/core/registry.py
42
+ src/superlocalmemory/core/summarizer.py
43
+ src/superlocalmemory/core/worker_pool.py
40
44
  src/superlocalmemory/dynamics/__init__.py
41
45
  src/superlocalmemory/dynamics/fisher_langevin_coupling.py
42
46
  src/superlocalmemory/encoding/__init__.py
package/ui/index.html CHANGED
@@ -1569,30 +1569,86 @@
1569
1569
  <h6 class="mb-0"><i class="bi bi-gear-wide-connected"></i> V3 Configuration</h6>
1570
1570
  </div>
1571
1571
  <div class="card-body">
1572
- <div class="row">
1573
- <div class="col-md-6">
1574
- <label class="form-label fw-bold">Operating Mode</label>
1575
- <select id="settings-mode" class="form-select">
1576
- <option value="a">Mode A — Local Guardian (Zero Cloud)</option>
1577
- <option value="b">Mode B — Smart Local (Ollama)</option>
1578
- <option value="c">Mode C — Full Power (Cloud LLM)</option>
1579
- </select>
1580
- <button class="btn btn-sm btn-primary mt-2" id="settings-mode-save">Save Mode</button>
1572
+ <!-- Current Mode Banner -->
1573
+ <div id="settings-current-banner" class="alert alert-info mb-3">
1574
+ <div class="d-flex justify-content-between align-items-center">
1575
+ <span>Active: <strong id="settings-current-mode">Loading...</strong></span>
1576
+ <span id="settings-current-detail" class="small"></span>
1581
1577
  </div>
1582
- <div class="col-md-6">
1583
- <label class="form-label fw-bold">LLM Provider</label>
1584
- <select id="settings-provider" class="form-select">
1585
- <option value="none">None (Mode A)</option>
1586
- <option value="openai">OpenAI</option>
1587
- <option value="anthropic">Anthropic</option>
1588
- <option value="ollama">Ollama (Local)</option>
1589
- <option value="openrouter">OpenRouter</option>
1590
- </select>
1591
- <div class="mt-2">
1592
- <input type="password" id="settings-api-key" class="form-control form-control-sm" placeholder="API Key">
1578
+ </div>
1579
+
1580
+ <!-- Step 1: Mode -->
1581
+ <div class="mb-3">
1582
+ <label class="form-label fw-bold">Step 1: Operating Mode</label>
1583
+ <div class="btn-group w-100" role="group">
1584
+ <input type="radio" class="btn-check" name="settings-mode-radio" id="mode-a-radio" value="a" checked>
1585
+ <label class="btn btn-outline-success" for="mode-a-radio">
1586
+ <strong>Mode A</strong><br><small>Zero Cloud — EU AI Act</small>
1587
+ </label>
1588
+ <input type="radio" class="btn-check" name="settings-mode-radio" id="mode-b-radio" value="b">
1589
+ <label class="btn btn-outline-info" for="mode-b-radio">
1590
+ <strong>Mode B</strong><br><small>Local Ollama LLM</small>
1591
+ </label>
1592
+ <input type="radio" class="btn-check" name="settings-mode-radio" id="mode-c-radio" value="c">
1593
+ <label class="btn btn-outline-warning" for="mode-c-radio">
1594
+ <strong>Mode C</strong><br><small>Cloud LLM (Best Accuracy)</small>
1595
+ </label>
1596
+ </div>
1597
+ </div>
1598
+
1599
+ <!-- Step 2: Provider Config (Mode B/C only) -->
1600
+ <div id="settings-provider-panel" style="display:none;" class="card p-3 mb-3 border-primary">
1601
+ <h6 class="mb-2">Step 2: LLM Configuration</h6>
1602
+
1603
+ <!-- Provider select -->
1604
+ <div class="row mb-2">
1605
+ <div class="col-md-4">
1606
+ <label class="form-label small fw-bold">Provider</label>
1607
+ <select id="settings-provider" class="form-select form-select-sm">
1608
+ <option value="">-- Select --</option>
1609
+ <option value="ollama">Ollama (Local)</option>
1610
+ <option value="openrouter">OpenRouter</option>
1611
+ <option value="openai">OpenAI</option>
1612
+ <option value="anthropic">Anthropic</option>
1613
+ </select>
1614
+ </div>
1615
+ <div class="col-md-4">
1616
+ <label class="form-label small fw-bold">Model <span class="text-danger">*</span></label>
1617
+ <select id="settings-model" class="form-select form-select-sm">
1618
+ <option value="">Select provider first</option>
1619
+ </select>
1620
+ <small id="settings-model-hint" class="text-muted"></small>
1621
+ </div>
1622
+ <div class="col-md-4" id="settings-key-col" style="display:none;">
1623
+ <label class="form-label small fw-bold">API Key <span class="text-danger">*</span></label>
1624
+ <input type="password" id="settings-api-key" class="form-control form-control-sm" placeholder="sk-... or your key">
1625
+ <small class="text-muted">Saved locally in ~/.superlocalmemory/</small>
1626
+ </div>
1627
+ </div>
1628
+
1629
+ <!-- Endpoint (advanced) -->
1630
+ <div class="row mb-2" id="settings-endpoint-row" style="display:none;">
1631
+ <div class="col-md-8">
1632
+ <label class="form-label small">API Endpoint (advanced)</label>
1633
+ <input type="text" id="settings-endpoint" class="form-control form-control-sm" placeholder="https://...">
1593
1634
  </div>
1594
- <button class="btn btn-sm btn-primary mt-2" id="settings-provider-save">Save Provider</button>
1595
1635
  </div>
1636
+
1637
+ <!-- Connection test -->
1638
+ <div>
1639
+ <button class="btn btn-sm btn-outline-primary" id="settings-test-btn">
1640
+ <i class="bi bi-lightning"></i> Test Connection
1641
+ </button>
1642
+ <span id="settings-test-result" class="ms-2 small"></span>
1643
+ </div>
1644
+ </div>
1645
+
1646
+ <!-- Save button -->
1647
+ <div class="mt-2">
1648
+ <button class="btn btn-primary" id="settings-save-all">
1649
+ <i class="bi bi-check-circle"></i> Save Configuration
1650
+ </button>
1651
+ <span id="settings-save-status" class="ms-2" style="display:none;"></span>
1596
1652
  </div>
1597
1653
  <hr>
1598
1654
  <div class="row">