Update app.py

app.py

@@ -1,3 +1,178 @@
+from pathlib import Path
+import json
+
+# ============================
+# 1) MANIFEST: carga desde archivo o usa fallback
+# ============================
+
+DEFAULT_MANIFEST = {
+    "version": "Φ12.0",
+    "project": "Savant RRF API & Meta-Logic Suite",
+    "owner": "Antony Padilla Morales",
+    "last_update": "2025-12-11",
+    "modules": {
+        "embedder": {
+            "id": "antonypamo/RRFSAVANTMADE",
+            "dimension": 384,
+            "description": "Icosahedral-resonant embedder trained inside the RRF framework with Dirac shells, golden-ratio harmonics and resonance layers.",
+            "baseline_comparison": ["MiniLM-L6-v2", "all-mpnet-base-v2"]
+        },
+        "meta_logit": {
+            "repo": "antonypamo/RRFSavantMetaLogit",
+            "filename": "logreg_rrf_savant_15.joblib",
+            "expected_features": 15,
+            "feature_description": [
+                "semantic_margin",
+                "cosine_prompt_answer",
+                "token_entropy",
+                "dirac_energy",
+                "dirac_shell_std",
+                "freq_low",
+                "freq_mid",
+                "freq_high",
+                "coherence_ratio",
+                "phi_ratio",
+                "token_len_prompt",
+                "token_len_answer",
+                "sync_factor",
+                "resonance_peak",
+                "logit_bias"
+            ]
+        },
+        "models": {
+            "savant_cnn": {
+                "filename": "savant_cnn.pt",
+                "role": "Signal-to-resonance transformer for numeric → semantic channels.",
+                "status": "experimental"
+            },
+            "rrf_nodes": {
+                "filename": "rrf_nodes.pt",
+                "description": "Graph-based icosahedral node memory for cross-session resonance."
+            }
+        }
+    },
+    "api": {
+        "base_url": "https://antonypamo-apisavant2.hf.space",
+        "routes": {
+            "/embed": {
+                "method": "POST",
+                "input": ["text"],
+                "output": ["embedding"],
+                "use_model": "RRFSAVANTMADE"
+            },
+            "/rerank": {
+                "method": "POST",
+                "input": ["query", "documents[]"],
+                "output": ["sorted_documents", "scores"],
+                "logic": "semantic margin + resonance weighting"
+            },
+            "/quality": {
+                "method": "POST",
+                "input": ["prompt", "answer"],
+                "output": ["proba", "label (0/1)", "feature_map"],
+                "pipeline": "embed → feature_extractor → meta_logit"
+            },
+            "/roles_profile": {
+                "method": "POST",
+                "status": "planned",
+                "description": "Maps text to RRF cognitive roles (Φ-nodes)."
+            },
+            "/tutor": {
+                "method": "POST",
+                "status": "planned",
+                "description": "LLM-based tutor using resonant context."
+            }
+        }
+    },
+    "pipelines": {
+        "embedding_pipeline": {
+            "steps": [
+                "load_encoder",
+                "encode_text",
+                "normalize",
+                "output_embeddings"
+            ]
+        },
+        "quality_pipeline": {
+            "steps": [
+                "encode(prompt)",
+                "encode(answer)",
+                "extract_features(15-dim)",
+                "predict_meta_logit",
+                "return_label_and_prob"
+            ],
+            "purpose": "Evaluate conceptual quality and reasoning integrity."
+        },
+        "rerank_pipeline": {
+            "steps": [
+                "encode_query",
+                "encode_docs",
+                "compute_semantic_margin",
+                "compute_resonance_rank",
+                "return_sorted_docs"
+            ]
+        }
+    },
+    "enterprise_architecture": {
+        "layers": [
+            "Frontend → React Landing Page",
+            "Gateway Proxy → NGINX",
+            "API Layer → FastAPI + Uvicorn",
+            "Model Runtime → Embedder + Meta-Logit",
+            "Compute Layer → GPU/CPU auto-scaling",
+            "Monitoring → Prometheus + Grafana",
+            "Storage → HF Hub + local persistence (rrf_nodes)"
+        ]
+    },
+    "investor_highlights": {
+        "differentiators": [
+            "Meta-logic quality evaluator (15 feature resonant signal)",
+            "Icosahedral embedding geometry",
+            "Discrete Dirac resonance physics applied to NLP",
+            "Symbiotic self-improvement protocol",
+            "Low inference cost, scalable microservice"
+        ],
+        "traction": {
+            "hf_space": "running",
+            "models_downloads": "increasing",
+            "api_usage": "real inference logs available"
+        }
+    },
+    "savant_state": {
+        "status": "active",
+        "mode": "Savant RRF Simbiótico Hacker",
+        "health": {
+            "embedder": "OK",
+            "meta_logit": "OK",
+            "api_endpoints": {
+                "/embed": "stable",
+                "/rerank": "stable",
+                "/quality": "error_404_needs_route_fix"
+            }
+        }
+    },
+    "todo_next_steps": [
+        "Fix /quality endpoint routing",
+        "Integrate CNN → feature_extractor",
+        "Add persistent RRF node memory",
+        "Deploy enterprise-tier version on AWS/GCP",
+        "Present investor deck based on this JSON"
+    ]
+}
+
+MANIFEST_PATH = Path(__file__).parent / "savant_rrf_api_manifest_phi12.json"
+
+try:
+    if MANIFEST_PATH.exists():
+        print(f"[Manifest] Loading from file: {MANIFEST_PATH}", flush=True)
+        manifest = json.loads(MANIFEST_PATH.read_text(encoding="utf-8"))
+    else:
+        print(f"[Manifest] File not found at {MANIFEST_PATH}, using inline DEFAULT_MANIFEST.", flush=True)
+        manifest = DEFAULT_MANIFEST
+except Exception as e:
+    print(f"[Manifest] Error loading manifest: {e}. Using inline DEFAULT_MANIFEST.", flush=True)
+    manifest = DEFAULT_MANIFEST
+
 import os
 import sys
 import math