Add TeleKnowledge vision to README - #5

…ion-and-readme-with-organizations-2025-07-29

Add TeleKnowledge vision to README

Added comprehensive Performance Benchmarks section (#5):
- Baseline sequential runs (WordNet: 145.87s, Fonts: ~780s, PDFs: 41.39s)
- Parallel optimized runs (1.02-3.6× speedup with 8 workers)
- Resource usage: <200MB VRAM, 6-8% GPU util (massive headroom!)
- Key findings: CPU-bound bottlenecks, 92-94% GPU idle

Added Sovereign Knowledge Ingestion Stack section:
- RPN-native embeddings (33,428 trigrams, 0MB footprint)
- Multi-modal pipeline architecture (text/audio/visual → Galaxy)
- Ingestion module reference
- Parallel optimization details

Updated Recent Milestones with Step 15 Phase B:
- 117,659 WordNet synsets + multi-lingual lexicons
- 168,206 font visual-text pairs
- 61 PDFs, 23,000 sentences ingested
- Zero external dependencies achieved

Real test results from Codex's tmux runs documented.

🤖 Generated with [Claude Code](https://claude.com/claude-code)

Co-Authored-By: Claude <noreply@anthropic.com>

…y-ctx-flags

Retry #5 reached get_global with cuMemGetInfo_v2 working but then hit
CUDA_ERROR_INVALID_CONTEXT (201) at cuModuleGetGlobal. Same v1/v2 symbol
bifurcation as 8a2b711: libcuda.so.1 exports cuModuleGetGlobal (v1,
0x2d67b0) and cuModuleGetGlobal_v2 (0x286d80) at distinct addresses.
Bare getattr on the ctypes CDLL resolves to the deprecated v1 symbol
which does not interoperate with v2-created contexts.

Extend the _bind_v2 resolver to:
  - _cuModuleGetGlobal          (v2 preferred)
  - _cuStreamDestroy            (v2 preferred, optional)
  - _cuDevicePrimaryCtxSetFlags (v2 preferred)

and route loader.get_global() / destroy_stream() / primary-ctx fallback
path through the explicit v2 pointers.

Verified: scripts/check_single_context_invariant.sh CLEAN; smoke bench
(sovereign_bitnet_attention.py --quick) now advances past get_global.

Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>