Spaces:
Runtime error
Runtime error
File size: 13,718 Bytes
edae06c 66b508d edae06c 66b508d edae06c 66b508d edae06c 66b508d edae06c 66b508d edae06c 66b508d edae06c 66b508d edae06c 66b508d edae06c 66b508d edae06c 66b508d edae06c 66b508d edae06c 66b508d edae06c 66b508d edae06c 66b508d edae06c 66b508d edae06c 66b508d edae06c 66b508d b460317 66b508d b460317 66b508d b460317 66b508d edae06c 66b508d |
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 |
"""
logos/ingest_knowledge.py - The "Token Consumption Plan"
Protocol 4: Autonomous Resource Integration -> Knowledge Synthesis
This script executes the "Ingestion" workflow:
1. Scans 'LOGOS Notes' for diagrammatic knowledge.
2. Uses the Nano Swarm (Gemma Vision) to "consume" the visual tokens.
3. Transmutes visual logic into textual knowledge (Markdown).
Target: 'LOGOS Notes/*.png' -> 'knowledge_base/diagram_analysis.md'
"""
import sys
import os
import glob
import time
from typing import List
sys.path.insert(0, os.path.abspath(os.path.join(os.path.dirname(__file__), '..')))
from logos.connectors import get_connector
# Configuration
SOURCE_DIR = "LOGOS Notes"
OUTPUT_FILE = "knowledge_base/diagram_analysis.md"
VISION_MODEL = "google/gemma-3-4b"
TEXT_MODEL = "dolphin-x1-8b"
def ingest_diagrams():
print(f"--- LOGOS Knowledge Ingestion [Protocol 4: Context-Primed Vision] ---")
print(f"Targeting: {SOURCE_DIR}")
print(f"Agents: {TEXT_MODEL} (Context) -> {VISION_MODEL} (Analysis)")
# 1. Scout Resources
images = glob.glob(os.path.join(SOURCE_DIR, "*.png")) + glob.glob(os.path.join(SOURCE_DIR, "*.jpg"))
images.sort()
if not images:
print(f"[WARN] No assets found in {SOURCE_DIR}")
return
# 2. Initialize Agents
try:
# We reuse the connector but switch 'model' param per call to stay hardware aligned (one active stream)
connector = get_connector('local')
except Exception as e:
print(f"[FAIL] Could not spawn connector: {e}")
return
# 3. Execution Loop
with open(OUTPUT_FILE, "w", encoding="utf-8") as f:
f.write("# LOGOS Diagram Analysis (Context-Primed)\n")
f.write(f"**Generated:** {time.strftime('%Y-%m-%d %H:%M:%S')}\n")
f.write(f"**Pipeline:** Context({TEXT_MODEL}) -> Vision({VISION_MODEL})\n\n")
for i, img_path in enumerate(images):
filename = os.path.basename(img_path)
clean_name = os.path.splitext(filename)[0].replace("_", " ").replace("-", " ")
print(f"[{i+1}/{len(images)}] Processing: {filename}...")
try:
# Force Hardware Cool-down between items
time.sleep(2.0)
# STEP A: Context Priming (Dolphin)
print(f" > [Dolphin] Extracting Context (timeout=30s)...")
try:
context_prompt = f"Analyze filename '{clean_name}'. Return JSON {{'title': '...', 'description': '...'}}."
# Debug: Print ensuring valid string
# print(f"DEBUG: Prompting Dolphin with: {context_prompt}")
context_resp, _ = connector.chat(context_prompt, model=TEXT_MODEL)
# Robust JSON extraction
import json
try:
# Find first '{' and last '}'
s = context_resp.find('{')
e = context_resp.rfind('}') + 1
if s != -1 and e != -1:
meta = json.loads(context_resp[s:e])
refined_context = f"Context: {meta.get('title', clean_name)}. {meta.get('description', '')}"
else:
refined_context = f"Context: {clean_name}"
except:
refined_context = f"Context: {context_resp[:200].replace(chr(10), ' ')}"
except Exception as e:
print(f" > [Dolphin] Bypass (Error: {e})")
refined_context = f"Context: {clean_name}"
# Hardware Switch Delay
print(f" > [System] Swapping to Vision Model...")
time.sleep(1.0)
# STEP B: Vision Analysis (Gemma)
print(f" > [Gemma] performing Context-Aware Analysis...")
vision_prompt = f"""
Role: Senior Hardware Architect.
Context: {refined_context}
Task: Analyze the diagram. Validate if the visual data matches the context.
Extract:
1. Hardware components (Bus, ALU, Cache).
2. Data flow direction.
3. Mathematical formulas.
Output: Actionable Markdown.
"""
start_ts = time.time()
analysis, _ = connector.chat(vision_prompt, image_path=img_path, model=VISION_MODEL)
duration = time.time() - start_ts
# Anneal into Knowledge Base
f.write(f"## {filename}\n")
f.write(f"**Context ({TEXT_MODEL}):** {refined_context}\n\n")
f.write(f"})\n\n")
f.write(f"{analysis}\n\n")
f.write(f"*Analysis time: {duration:.2f}s*\n")
f.write("---\n\n")
f.flush()
print(f" > Ingested ({duration:.2f}s)")
except Exception as e:
print(f" > [FAIL] Error: {e}")
f.write(f"## {filename}\n")
f.write(f"**Error:** {e}\n\n---\n\n")
print(f"\n[SUCCESS] Knowledge synthesis complete.")
print(f"Artifact: {OUTPUT_FILE}")
# --- INGESTION REGISTRY (Deduplication) ---
class IngestionRegistry:
def __init__(self, registry_path="logos/ingestion_registry.json"):
self.registry_path = registry_path
self.data = self._load()
def _load(self):
if os.path.exists(self.registry_path):
try:
import json
with open(self.registry_path, 'r') as f:
return json.load(f)
except:
pass
return {}
def save(self):
import json
with open(self.registry_path, 'w') as f:
json.dump(self.data, f, indent=2)
def is_processed(self, filepath):
"""Checks if file is already ingested based on mtime."""
stat = os.stat(filepath)
key = os.path.abspath(filepath)
last_mtime = self.data.get(key, {}).get("mtime", 0)
return stat.st_mtime <= last_mtime
def mark_processed(self, filepath, meta=None):
"""Tags data as ingested."""
key = os.path.abspath(filepath)
self.data[key] = {
"mtime": os.stat(filepath).st_mtime,
"timestamp": time.time(),
"meta": meta or {}
}
self.save()
def ingest_diagrams():
print(f"--- LOGOS Knowledge Ingestion [Protocol 4: Context-Primed Vision] ---")
print(f"Targeting: {SOURCE_DIR}")
print(f"Agents: {TEXT_MODEL} (Context) -> {VISION_MODEL} (Analysis)")
registry = IngestionRegistry()
# 1. Scout Resources
images = glob.glob(os.path.join(SOURCE_DIR, "*.png")) + glob.glob(os.path.join(SOURCE_DIR, "*.jpg"))
images.sort()
if not images:
print(f"[WARN] No assets found in {SOURCE_DIR}")
return
# 2. Initialize Agents
try:
# We reuse the connector but switch 'model' param per call to stay hardware aligned (one active stream)
connector = get_connector('local')
except Exception as e:
print(f"[FAIL] Could not spawn connector: {e}")
return
# 3. Execution Loop
# Write mode 'a' (append) to preserve history if we are skipping?
# Or 'w' but reading old content?
# For now, let's just append new findings or overwrite if it's a full run.
# User asked to "tag data that we have already ingested", usually implies skipping.
# Check if output file exists, if so append, else write header
mode = 'a' if os.path.exists(OUTPUT_FILE) else 'w'
with open(OUTPUT_FILE, mode, encoding="utf-8") as f:
if mode == 'w':
f.write("# LOGOS Diagram Analysis (Context-Primed)\n")
f.write(f"**Generated:** {time.strftime('%Y-%m-%d %H:%M:%S')}\n")
f.write(f"**Pipeline:** Context({TEXT_MODEL}) -> Vision({VISION_MODEL})\n\n")
for i, img_path in enumerate(images):
filename = os.path.basename(img_path)
# [DEDUPLICATION CHECK]
if registry.is_processed(img_path):
print(f"[{i+1}/{len(images)}] Skipping {filename} (Already Ingested) [Duplicate Path Atom]")
continue
clean_name = os.path.splitext(filename)[0].replace("_", " ").replace("-", " ")
print(f"[{i+1}/{len(images)}] Processing: {filename}...")
try:
# Force Hardware Cool-down between items
time.sleep(1.0)
# STEP A: Context Priming (Dolphin)
print(f" > [Dolphin] Extracting Context (timeout=30s)...")
try:
context_prompt = f"Analyze filename '{clean_name}'. Return JSON {{'title': '...', 'description': '...'}}."
context_resp, _ = connector.chat(context_prompt, model=TEXT_MODEL)
# Robust JSON extraction
import json
try:
s = context_resp.find('{')
e = context_resp.rfind('}') + 1
if s != -1 and e != -1:
meta = json.loads(context_resp[s:e])
refined_context = f"Context: {meta.get('title', clean_name)}. {meta.get('description', '')}"
else:
refined_context = f"Context: {clean_name}"
except:
refined_context = f"Context: {context_resp[:200].replace(chr(10), ' ')}"
except Exception as e:
print(f" > [Dolphin] Bypass (Error: {e})")
refined_context = f"Context: {clean_name}"
# Hardware Switch Delay
print(f" > [System] Swapping to Vision Model...")
time.sleep(1.0)
# STEP B: Vision Analysis (Gemma)
print(f" > [Gemma] performing Context-Aware Analysis...")
# [CACHE OPTIMIZATION]
# We move the Static Instructions to the FRONT and the Dynamic Context to the BACK.
# Dynamic part
user_content = f"Context: {refined_context}\nTarget: {filename}"
# Static System/Instruction Block (Cached)
vision_system = (
"Role: Senior Hardware Architect.\n"
"Task: Analyze the diagram. Validate if the visual data matches the context.\n"
"Extract:\n"
"1. Hardware components (Bus, ALU, Cache).\n"
"2. Data flow direction.\n"
"3. Mathematical formulas.\n"
"Output: Actionable Markdown."
)
full_prompt = f"{vision_system}\n\n[DATA]: {user_content}"
start_ts = time.time()
analysis, _ = connector.chat(full_prompt, image_path=img_path, model=VISION_MODEL)
duration = time.time() - start_ts
# Anneal into Knowledge Base
f.write(f"## {filename}\n")
f.write(f"**Context ({TEXT_MODEL}):** {refined_context}\n\n")
f.write(f"})\n\n")
f.write(f"{analysis}\n\n")
f.write(f"*Analysis time: {duration:.2f}s*\n")
f.write("---\n\n")
f.flush()
print(f" > Ingested ({duration:.2f}s)")
# [PROTOCOL 26: ANALOG TO DIGITAL INDEXING]
# We send the Vision Analysis to the Server to be converted into a Gödel Number
try:
import requests
payload = {
"filepath": img_path,
"content": f"{refined_context}\n{analysis}"
}
# Assuming default port
res = requests.post("http://localhost:5000/index-module", json=payload, timeout=5)
if res.status_code == 200:
data = res.json()
mid = data.get('manifold_id')
primes = data.get('prime_coordinates')
print(f" > [GÖDEL] Indexing Complete. Manifold ID: {mid}")
print(f" > [PRIME SPACE] Active Domains: {primes}")
f.write(f"**Gödel ID:** `{mid}`\n**Prime Vectors:** `{primes}`\n\n")
else:
print(f" > [GÖDEL] Indexing Failed: {res.status_code}")
except Exception as e:
print(f" > [GÖDEL] Indexing Error: {e}")
# [REGISTRY UPDATE]
registry.mark_processed(img_path, meta={"duration": duration, "context": refined_context})
except Exception as e:
print(f" > [FAIL] Error: {e}")
f.write(f"## {filename}\n")
f.write(f"**Error:** {e}\n\n---\n\n")
print(f"\n[SUCCESS] Knowledge synthesis complete.")
if __name__ == "__main__":
ingest_diagrams()
ingest_documents()
|