awareness-memory-cloud 2.4.3__tar.gz → 2.4.4__tar.gz

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.
Files changed (37) hide show
  1. {awareness_memory_cloud-2.4.3 → awareness_memory_cloud-2.4.4}/PKG-INFO +90 -3
  2. {awareness_memory_cloud-2.4.3 → awareness_memory_cloud-2.4.4}/README.md +89 -2
  3. {awareness_memory_cloud-2.4.3 → awareness_memory_cloud-2.4.4}/awareness_memory_cloud.egg-info/PKG-INFO +90 -3
  4. {awareness_memory_cloud-2.4.3 → awareness_memory_cloud-2.4.4}/pyproject.toml +1 -1
  5. {awareness_memory_cloud-2.4.3 → awareness_memory_cloud-2.4.4}/LICENSE +0 -0
  6. {awareness_memory_cloud-2.4.3 → awareness_memory_cloud-2.4.4}/awareness_memory_cloud.egg-info/SOURCES.txt +0 -0
  7. {awareness_memory_cloud-2.4.3 → awareness_memory_cloud-2.4.4}/awareness_memory_cloud.egg-info/dependency_links.txt +0 -0
  8. {awareness_memory_cloud-2.4.3 → awareness_memory_cloud-2.4.4}/awareness_memory_cloud.egg-info/requires.txt +0 -0
  9. {awareness_memory_cloud-2.4.3 → awareness_memory_cloud-2.4.4}/awareness_memory_cloud.egg-info/top_level.txt +0 -0
  10. {awareness_memory_cloud-2.4.3 → awareness_memory_cloud-2.4.4}/memory_cloud/__init__.py +0 -0
  11. {awareness_memory_cloud-2.4.3 → awareness_memory_cloud-2.4.4}/memory_cloud/client.py +0 -0
  12. {awareness_memory_cloud-2.4.3 → awareness_memory_cloud-2.4.4}/memory_cloud/errors.py +0 -0
  13. {awareness_memory_cloud-2.4.3 → awareness_memory_cloud-2.4.4}/memory_cloud/export_reader.py +0 -0
  14. {awareness_memory_cloud-2.4.3 → awareness_memory_cloud-2.4.4}/memory_cloud/integrations/__init__.py +0 -0
  15. {awareness_memory_cloud-2.4.3 → awareness_memory_cloud-2.4.4}/memory_cloud/integrations/_base.py +0 -0
  16. {awareness_memory_cloud-2.4.3 → awareness_memory_cloud-2.4.4}/memory_cloud/integrations/autogen.py +0 -0
  17. {awareness_memory_cloud-2.4.3 → awareness_memory_cloud-2.4.4}/memory_cloud/integrations/crewai.py +0 -0
  18. {awareness_memory_cloud-2.4.3 → awareness_memory_cloud-2.4.4}/memory_cloud/integrations/langchain.py +0 -0
  19. {awareness_memory_cloud-2.4.3 → awareness_memory_cloud-2.4.4}/memory_cloud/integrations/praisonai.py +0 -0
  20. {awareness_memory_cloud-2.4.3 → awareness_memory_cloud-2.4.4}/memory_cloud/interceptor.py +0 -0
  21. {awareness_memory_cloud-2.4.3 → awareness_memory_cloud-2.4.4}/memory_cloud/query_rewrite.py +0 -0
  22. {awareness_memory_cloud-2.4.3 → awareness_memory_cloud-2.4.4}/memory_cloud/quickstart.py +0 -0
  23. {awareness_memory_cloud-2.4.3 → awareness_memory_cloud-2.4.4}/memory_cloud/types.py +0 -0
  24. {awareness_memory_cloud-2.4.3 → awareness_memory_cloud-2.4.4}/setup.cfg +0 -0
  25. {awareness_memory_cloud-2.4.3 → awareness_memory_cloud-2.4.4}/tests/test_client_chat_and_record.py +0 -0
  26. {awareness_memory_cloud-2.4.3 → awareness_memory_cloud-2.4.4}/tests/test_client_crud.py +0 -0
  27. {awareness_memory_cloud-2.4.3 → awareness_memory_cloud-2.4.4}/tests/test_client_extraction_passthrough.py +0 -0
  28. {awareness_memory_cloud-2.4.3 → awareness_memory_cloud-2.4.4}/tests/test_client_recall_and_compaction.py +0 -0
  29. {awareness_memory_cloud-2.4.3 → awareness_memory_cloud-2.4.4}/tests/test_client_tasks_agents_users.py +0 -0
  30. {awareness_memory_cloud-2.4.3 → awareness_memory_cloud-2.4.4}/tests/test_client_upload_export_keys.py +0 -0
  31. {awareness_memory_cloud-2.4.3 → awareness_memory_cloud-2.4.4}/tests/test_ecosystem_smoke.py +0 -0
  32. {awareness_memory_cloud-2.4.3 → awareness_memory_cloud-2.4.4}/tests/test_export_reader.py +0 -0
  33. {awareness_memory_cloud-2.4.3 → awareness_memory_cloud-2.4.4}/tests/test_integrations.py +0 -0
  34. {awareness_memory_cloud-2.4.3 → awareness_memory_cloud-2.4.4}/tests/test_interceptor_response_format_fallback.py +0 -0
  35. {awareness_memory_cloud-2.4.3 → awareness_memory_cloud-2.4.4}/tests/test_interceptor_wrap.py +0 -0
  36. {awareness_memory_cloud-2.4.3 → awareness_memory_cloud-2.4.4}/tests/test_quickstart.py +0 -0
  37. {awareness_memory_cloud-2.4.3 → awareness_memory_cloud-2.4.4}/tests/test_sdk_functional.py +0 -0
{awareness_memory_cloud-2.4.3 → awareness_memory_cloud-2.4.4}/PKG-INFO RENAMED
@@ -1,6 +1,6 @@
1
1
  Metadata-Version: 2.4
2
2
  Name: awareness-memory-cloud
3
- Version: 2.4.3
3
+ Version: 2.4.4
4
4
  Summary: Python SDK for Awareness Memory Cloud
5
5
  Author: Awareness Team
6
6
  License: Apache License
@@ -249,12 +249,99 @@ Dynamic: license-file
249
249
 
250
250
  # Awareness Memory SDK — Python
251
251
 
252
- [![PyPI](https://img.shields.io/pypi/v/awareness-memory-cloud?color=00d4ff)](https://pypi.org/project/awareness-memory-cloud/) [![Discord](https://img.shields.io/discord/1354000000000000000?color=5865F2&logo=discord&logoColor=white&label=Discord)](https://discord.com/invite/nMDrT538Qa)
252
+ [![PyPI](https://img.shields.io/pypi/v/awareness-memory-cloud?color=00d4ff)](https://pypi.org/project/awareness-memory-cloud/) [![LongMemEval R@5](https://img.shields.io/badge/LongMemEval_R%405-95.6%25-brightgreen)](https://arxiv.org/abs/2410.10813) [![Discord](https://img.shields.io/discord/1354000000000000000?color=5865F2&logo=discord&logoColor=white&label=Discord)](https://discord.com/invite/nMDrT538Qa)
253
253
 
254
- Python SDK for adding persistent memory to AI agents and apps.
254
+ Python SDK for adding persistent memory to AI agents and apps. **95.6% Recall@5 on [LongMemEval](https://arxiv.org/abs/2410.10813) (ICLR 2025)**.
255
255
 
256
256
  Online docs: <https://awareness.market/docs?doc=python>
257
257
 
258
+ ## Benchmark: LongMemEval (ICLR 2025)
259
+
260
+ Awareness Memory is evaluated on **[LongMemEval](https://arxiv.org/abs/2410.10813)** — the industry standard benchmark for long-term conversational memory, published at ICLR 2025. 500 human-curated questions across 5 core capabilities.
261
+
262
+ ```
263
+ ╔══════════════════════════════════════════════════════════════╗
264
+ ║ ║
265
+ ║ Awareness Memory — LongMemEval Benchmark Results ║
266
+ ║ ───────────────────────────────────────────────── ║
267
+ ║ ║
268
+ ║ Benchmark: LongMemEval (ICLR 2025) ║
269
+ ║ Dataset: 500 human-curated questions ║
270
+ ║ Variant: LongMemEval_S (~115k tokens per question) ║
271
+ ║ ║
272
+ ║ ┌─────────────────────────────────────────────────┐ ║
273
+ ║ │ │ ║
274
+ ║ │ Recall@1 77.6% (388 / 500) │ ║
275
+ ║ │ Recall@3 91.8% (459 / 500) │ ║
276
+ ║ │ Recall@5 95.6% (478 / 500) ◀ PRIMARY │ ║
277
+ ║ │ Recall@10 97.4% (487 / 500) │ ║
278
+ ║ │ │ ║
279
+ ║ └─────────────────────────────────────────────────┘ ║
280
+ ║ ║
281
+ ║ Method: Hybrid RRF (BM25 + Semantic Vector Search) ║
282
+ ║ Embedding: all-MiniLM-L6-v2 (384d) ║
283
+ ║ LLM Calls: 0 (pure retrieval, no generation cost) ║
284
+ ║ Hardware: Apple M1, 8GB RAM — 14 min total ║
285
+ ║ ║
286
+ ╚══════════════════════════════════════════════════════════════╝
287
+ ```
288
+
289
+ ### Leaderboard
290
+
291
+ ```
292
+ ┌─────────────────────────────────────────────────────────────┐
293
+ │ Long-Term Memory Retrieval — R@5 Leaderboard │
294
+ │ LongMemEval (ICLR 2025, 500 questions) │
295
+ ├─────────────────────────────────┬───────────┬───────────────┤
296
+ │ System │ R@5 │ Note │
297
+ ├─────────────────────────────────┼───────────┼───────────────┤
298
+ │ MemPalace (ChromaDB raw) │ 96.6% │ R@5 only * │
299
+ │ ★ Awareness Memory (Hybrid) │ 95.6% │ Hybrid RRF │
300
+ │ OMEGA │ 95.4% │ QA Accuracy │
301
+ │ Mastra (GPT-5-mini) │ 94.9% │ QA Accuracy │
302
+ │ Mastra (GPT-4o) │ 84.2% │ QA Accuracy │
303
+ │ Supermemory │ 81.6% │ QA Accuracy │
304
+ │ Zep / Graphiti │ 71.2% │ QA Accuracy │
305
+ │ GPT-4o (full context) │ 60.6% │ QA Accuracy │
306
+ ├─────────────────────────────────┴───────────┴───────────────┤
307
+ │ * MemPalace 96.6% is Recall@5 only, not QA Accuracy. │
308
+ │ Palace hierarchy was NOT used in the evaluation. │
309
+ └─────────────────────────────────────────────────────────────┘
310
+ ```
311
+
312
+ ### Accuracy by Question Type
313
+
314
+ ```
315
+ ┌─────────────────────────────────────────────────────────────┐
316
+ │ Awareness Memory — R@5 by Question Type │
317
+ │ │
318
+ │ knowledge-update ████████████████████████████ 100% │
319
+ │ multi-session ███████████████████████████▋ 98.5%│
320
+ │ single-session-asst ███████████████████████████▌ 98.2%│
321
+ │ temporal-reasoning █████████████████████████▊ 94.7%│
322
+ │ single-session-user ████████████████████████▎ 88.6%│
323
+ │ single-session-pref ███████████████████████▏ 86.7%│
324
+ │ │
325
+ │ Overall █████████████████████████▉ 95.6%│
326
+ │ │
327
+ │ ┌───────────────────────────────────────────────┐ │
328
+ │ │ Ablation Study │ │
329
+ │ │ ───────────────────────────────────────── │ │
330
+ │ │ Vector-only: 92.6% ▓▓▓▓▓▓▓▓▓▓▓▓▓░░░ │ │
331
+ │ │ BM25-only: 91.4% ▓▓▓▓▓▓▓▓▓▓▓▓▓░░░ │ │
332
+ │ │ Hybrid RRF: 95.6% ▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓░ ★ │ │
333
+ │ │ Hybrid = +3% over any │ │
334
+ │ │ single method alone │ │
335
+ │ └───────────────────────────────────────────────┘ │
336
+ │ │
337
+ │ arxiv.org/abs/2410.10813 awareness.market │
338
+ └─────────────────────────────────────────────────────────────┘
339
+ ```
340
+
341
+ Zero LLM calls. Runs on Apple M1 8GB in 14 minutes. [Reproducible benchmark scripts →](https://github.com/edwin-hao-ai/Awareness/tree/main/benchmarks/longmemeval)
342
+
343
+ ---
344
+
258
345
  ## Install
259
346
 
260
347
  ```bash
{awareness_memory_cloud-2.4.3 → awareness_memory_cloud-2.4.4}/README.md RENAMED
@@ -1,11 +1,98 @@
1
1
  # Awareness Memory SDK — Python
2
2
 
3
- [![PyPI](https://img.shields.io/pypi/v/awareness-memory-cloud?color=00d4ff)](https://pypi.org/project/awareness-memory-cloud/) [![Discord](https://img.shields.io/discord/1354000000000000000?color=5865F2&logo=discord&logoColor=white&label=Discord)](https://discord.com/invite/nMDrT538Qa)
3
+ [![PyPI](https://img.shields.io/pypi/v/awareness-memory-cloud?color=00d4ff)](https://pypi.org/project/awareness-memory-cloud/) [![LongMemEval R@5](https://img.shields.io/badge/LongMemEval_R%405-95.6%25-brightgreen)](https://arxiv.org/abs/2410.10813) [![Discord](https://img.shields.io/discord/1354000000000000000?color=5865F2&logo=discord&logoColor=white&label=Discord)](https://discord.com/invite/nMDrT538Qa)
4
4
 
5
- Python SDK for adding persistent memory to AI agents and apps.
5
+ Python SDK for adding persistent memory to AI agents and apps. **95.6% Recall@5 on [LongMemEval](https://arxiv.org/abs/2410.10813) (ICLR 2025)**.
6
6
 
7
7
  Online docs: <https://awareness.market/docs?doc=python>
8
8
 
9
+ ## Benchmark: LongMemEval (ICLR 2025)
10
+
11
+ Awareness Memory is evaluated on **[LongMemEval](https://arxiv.org/abs/2410.10813)** — the industry standard benchmark for long-term conversational memory, published at ICLR 2025. 500 human-curated questions across 5 core capabilities.
12
+
13
+ ```
14
+ ╔══════════════════════════════════════════════════════════════╗
15
+ ║ ║
16
+ ║ Awareness Memory — LongMemEval Benchmark Results ║
17
+ ║ ───────────────────────────────────────────────── ║
18
+ ║ ║
19
+ ║ Benchmark: LongMemEval (ICLR 2025) ║
20
+ ║ Dataset: 500 human-curated questions ║
21
+ ║ Variant: LongMemEval_S (~115k tokens per question) ║
22
+ ║ ║
23
+ ║ ┌─────────────────────────────────────────────────┐ ║
24
+ ║ │ │ ║
25
+ ║ │ Recall@1 77.6% (388 / 500) │ ║
26
+ ║ │ Recall@3 91.8% (459 / 500) │ ║
27
+ ║ │ Recall@5 95.6% (478 / 500) ◀ PRIMARY │ ║
28
+ ║ │ Recall@10 97.4% (487 / 500) │ ║
29
+ ║ │ │ ║
30
+ ║ └─────────────────────────────────────────────────┘ ║
31
+ ║ ║
32
+ ║ Method: Hybrid RRF (BM25 + Semantic Vector Search) ║
33
+ ║ Embedding: all-MiniLM-L6-v2 (384d) ║
34
+ ║ LLM Calls: 0 (pure retrieval, no generation cost) ║
35
+ ║ Hardware: Apple M1, 8GB RAM — 14 min total ║
36
+ ║ ║
37
+ ╚══════════════════════════════════════════════════════════════╝
38
+ ```
39
+
40
+ ### Leaderboard
41
+
42
+ ```
43
+ ┌─────────────────────────────────────────────────────────────┐
44
+ │ Long-Term Memory Retrieval — R@5 Leaderboard │
45
+ │ LongMemEval (ICLR 2025, 500 questions) │
46
+ ├─────────────────────────────────┬───────────┬───────────────┤
47
+ │ System │ R@5 │ Note │
48
+ ├─────────────────────────────────┼───────────┼───────────────┤
49
+ │ MemPalace (ChromaDB raw) │ 96.6% │ R@5 only * │
50
+ │ ★ Awareness Memory (Hybrid) │ 95.6% │ Hybrid RRF │
51
+ │ OMEGA │ 95.4% │ QA Accuracy │
52
+ │ Mastra (GPT-5-mini) │ 94.9% │ QA Accuracy │
53
+ │ Mastra (GPT-4o) │ 84.2% │ QA Accuracy │
54
+ │ Supermemory │ 81.6% │ QA Accuracy │
55
+ │ Zep / Graphiti │ 71.2% │ QA Accuracy │
56
+ │ GPT-4o (full context) │ 60.6% │ QA Accuracy │
57
+ ├─────────────────────────────────┴───────────┴───────────────┤
58
+ │ * MemPalace 96.6% is Recall@5 only, not QA Accuracy. │
59
+ │ Palace hierarchy was NOT used in the evaluation. │
60
+ └─────────────────────────────────────────────────────────────┘
61
+ ```
62
+
63
+ ### Accuracy by Question Type
64
+
65
+ ```
66
+ ┌─────────────────────────────────────────────────────────────┐
67
+ │ Awareness Memory — R@5 by Question Type │
68
+ │ │
69
+ │ knowledge-update ████████████████████████████ 100% │
70
+ │ multi-session ███████████████████████████▋ 98.5%│
71
+ │ single-session-asst ███████████████████████████▌ 98.2%│
72
+ │ temporal-reasoning █████████████████████████▊ 94.7%│
73
+ │ single-session-user ████████████████████████▎ 88.6%│
74
+ │ single-session-pref ███████████████████████▏ 86.7%│
75
+ │ │
76
+ │ Overall █████████████████████████▉ 95.6%│
77
+ │ │
78
+ │ ┌───────────────────────────────────────────────┐ │
79
+ │ │ Ablation Study │ │
80
+ │ │ ───────────────────────────────────────── │ │
81
+ │ │ Vector-only: 92.6% ▓▓▓▓▓▓▓▓▓▓▓▓▓░░░ │ │
82
+ │ │ BM25-only: 91.4% ▓▓▓▓▓▓▓▓▓▓▓▓▓░░░ │ │
83
+ │ │ Hybrid RRF: 95.6% ▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓░ ★ │ │
84
+ │ │ Hybrid = +3% over any │ │
85
+ │ │ single method alone │ │
86
+ │ └───────────────────────────────────────────────┘ │
87
+ │ │
88
+ │ arxiv.org/abs/2410.10813 awareness.market │
89
+ └─────────────────────────────────────────────────────────────┘
90
+ ```
91
+
92
+ Zero LLM calls. Runs on Apple M1 8GB in 14 minutes. [Reproducible benchmark scripts →](https://github.com/edwin-hao-ai/Awareness/tree/main/benchmarks/longmemeval)
93
+
94
+ ---
95
+
9
96
  ## Install
10
97
 
11
98
  ```bash
{awareness_memory_cloud-2.4.3 → awareness_memory_cloud-2.4.4}/awareness_memory_cloud.egg-info/PKG-INFO RENAMED
@@ -1,6 +1,6 @@
1
1
  Metadata-Version: 2.4
2
2
  Name: awareness-memory-cloud
3
- Version: 2.4.3
3
+ Version: 2.4.4
4
4
  Summary: Python SDK for Awareness Memory Cloud
5
5
  Author: Awareness Team
6
6
  License: Apache License
@@ -249,12 +249,99 @@ Dynamic: license-file
249
249
 
250
250
  # Awareness Memory SDK — Python
251
251
 
252
- [![PyPI](https://img.shields.io/pypi/v/awareness-memory-cloud?color=00d4ff)](https://pypi.org/project/awareness-memory-cloud/) [![Discord](https://img.shields.io/discord/1354000000000000000?color=5865F2&logo=discord&logoColor=white&label=Discord)](https://discord.com/invite/nMDrT538Qa)
252
+ [![PyPI](https://img.shields.io/pypi/v/awareness-memory-cloud?color=00d4ff)](https://pypi.org/project/awareness-memory-cloud/) [![LongMemEval R@5](https://img.shields.io/badge/LongMemEval_R%405-95.6%25-brightgreen)](https://arxiv.org/abs/2410.10813) [![Discord](https://img.shields.io/discord/1354000000000000000?color=5865F2&logo=discord&logoColor=white&label=Discord)](https://discord.com/invite/nMDrT538Qa)
253
253
 
254
- Python SDK for adding persistent memory to AI agents and apps.
254
+ Python SDK for adding persistent memory to AI agents and apps. **95.6% Recall@5 on [LongMemEval](https://arxiv.org/abs/2410.10813) (ICLR 2025)**.
255
255
 
256
256
  Online docs: <https://awareness.market/docs?doc=python>
257
257
 
258
+ ## Benchmark: LongMemEval (ICLR 2025)
259
+
260
+ Awareness Memory is evaluated on **[LongMemEval](https://arxiv.org/abs/2410.10813)** — the industry standard benchmark for long-term conversational memory, published at ICLR 2025. 500 human-curated questions across 5 core capabilities.
261
+
262
+ ```
263
+ ╔══════════════════════════════════════════════════════════════╗
264
+ ║ ║
265
+ ║ Awareness Memory — LongMemEval Benchmark Results ║
266
+ ║ ───────────────────────────────────────────────── ║
267
+ ║ ║
268
+ ║ Benchmark: LongMemEval (ICLR 2025) ║
269
+ ║ Dataset: 500 human-curated questions ║
270
+ ║ Variant: LongMemEval_S (~115k tokens per question) ║
271
+ ║ ║
272
+ ║ ┌─────────────────────────────────────────────────┐ ║
273
+ ║ │ │ ║
274
+ ║ │ Recall@1 77.6% (388 / 500) │ ║
275
+ ║ │ Recall@3 91.8% (459 / 500) │ ║
276
+ ║ │ Recall@5 95.6% (478 / 500) ◀ PRIMARY │ ║
277
+ ║ │ Recall@10 97.4% (487 / 500) │ ║
278
+ ║ │ │ ║
279
+ ║ └─────────────────────────────────────────────────┘ ║
280
+ ║ ║
281
+ ║ Method: Hybrid RRF (BM25 + Semantic Vector Search) ║
282
+ ║ Embedding: all-MiniLM-L6-v2 (384d) ║
283
+ ║ LLM Calls: 0 (pure retrieval, no generation cost) ║
284
+ ║ Hardware: Apple M1, 8GB RAM — 14 min total ║
285
+ ║ ║
286
+ ╚══════════════════════════════════════════════════════════════╝
287
+ ```
288
+
289
+ ### Leaderboard
290
+
291
+ ```
292
+ ┌─────────────────────────────────────────────────────────────┐
293
+ │ Long-Term Memory Retrieval — R@5 Leaderboard │
294
+ │ LongMemEval (ICLR 2025, 500 questions) │
295
+ ├─────────────────────────────────┬───────────┬───────────────┤
296
+ │ System │ R@5 │ Note │
297
+ ├─────────────────────────────────┼───────────┼───────────────┤
298
+ │ MemPalace (ChromaDB raw) │ 96.6% │ R@5 only * │
299
+ │ ★ Awareness Memory (Hybrid) │ 95.6% │ Hybrid RRF │
300
+ │ OMEGA │ 95.4% │ QA Accuracy │
301
+ │ Mastra (GPT-5-mini) │ 94.9% │ QA Accuracy │
302
+ │ Mastra (GPT-4o) │ 84.2% │ QA Accuracy │
303
+ │ Supermemory │ 81.6% │ QA Accuracy │
304
+ │ Zep / Graphiti │ 71.2% │ QA Accuracy │
305
+ │ GPT-4o (full context) │ 60.6% │ QA Accuracy │
306
+ ├─────────────────────────────────┴───────────┴───────────────┤
307
+ │ * MemPalace 96.6% is Recall@5 only, not QA Accuracy. │
308
+ │ Palace hierarchy was NOT used in the evaluation. │
309
+ └─────────────────────────────────────────────────────────────┘
310
+ ```
311
+
312
+ ### Accuracy by Question Type
313
+
314
+ ```
315
+ ┌─────────────────────────────────────────────────────────────┐
316
+ │ Awareness Memory — R@5 by Question Type │
317
+ │ │
318
+ │ knowledge-update ████████████████████████████ 100% │
319
+ │ multi-session ███████████████████████████▋ 98.5%│
320
+ │ single-session-asst ███████████████████████████▌ 98.2%│
321
+ │ temporal-reasoning █████████████████████████▊ 94.7%│
322
+ │ single-session-user ████████████████████████▎ 88.6%│
323
+ │ single-session-pref ███████████████████████▏ 86.7%│
324
+ │ │
325
+ │ Overall █████████████████████████▉ 95.6%│
326
+ │ │
327
+ │ ┌───────────────────────────────────────────────┐ │
328
+ │ │ Ablation Study │ │
329
+ │ │ ───────────────────────────────────────── │ │
330
+ │ │ Vector-only: 92.6% ▓▓▓▓▓▓▓▓▓▓▓▓▓░░░ │ │
331
+ │ │ BM25-only: 91.4% ▓▓▓▓▓▓▓▓▓▓▓▓▓░░░ │ │
332
+ │ │ Hybrid RRF: 95.6% ▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓░ ★ │ │
333
+ │ │ Hybrid = +3% over any │ │
334
+ │ │ single method alone │ │
335
+ │ └───────────────────────────────────────────────┘ │
336
+ │ │
337
+ │ arxiv.org/abs/2410.10813 awareness.market │
338
+ └─────────────────────────────────────────────────────────────┘
339
+ ```
340
+
341
+ Zero LLM calls. Runs on Apple M1 8GB in 14 minutes. [Reproducible benchmark scripts →](https://github.com/edwin-hao-ai/Awareness/tree/main/benchmarks/longmemeval)
342
+
343
+ ---
344
+
258
345
  ## Install
259
346
 
260
347
  ```bash
{awareness_memory_cloud-2.4.3 → awareness_memory_cloud-2.4.4}/pyproject.toml RENAMED
@@ -4,7 +4,7 @@ build-backend = "setuptools.build_meta"
4
4
 
5
5
  [project]
6
6
  name = "awareness-memory-cloud"
7
- version = "2.4.3"
7
+ version = "2.4.4"
8
8
  description = "Python SDK for Awareness Memory Cloud"
9
9
  readme = "README.md"
10
10
  license = { file = "LICENSE" }