aither-kvcache 2.4.1__tar.gz → 2.4.2__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.
- {aither_kvcache-2.4.1 → aither_kvcache-2.4.2}/PKG-INFO +22 -12
- {aither_kvcache-2.4.1 → aither_kvcache-2.4.2}/README.md +21 -11
- {aither_kvcache-2.4.1 → aither_kvcache-2.4.2}/aither_kvcache/__init__.py +8 -1
- aither_kvcache-2.4.2/aither_kvcache/kvtransfer/align.py +448 -0
- {aither_kvcache-2.4.1 → aither_kvcache-2.4.2}/aither_kvcache/kvtransfer/capture.py +213 -5
- {aither_kvcache-2.4.1 → aither_kvcache-2.4.2}/aither_kvcache/kvtransfer/fit.py +56 -2
- {aither_kvcache-2.4.1 → aither_kvcache-2.4.2}/aither_kvcache/kvtransfer/geometry.py +43 -5
- {aither_kvcache-2.4.1 → aither_kvcache-2.4.2}/aither_kvcache/kvtransfer/pack.py +134 -4
- {aither_kvcache-2.4.1 → aither_kvcache-2.4.2}/aither_kvcache/kvtransfer/transfer.py +162 -15
- {aither_kvcache-2.4.1 → aither_kvcache-2.4.2}/aither_kvcache.egg-info/PKG-INFO +22 -12
- {aither_kvcache-2.4.1 → aither_kvcache-2.4.2}/aither_kvcache.egg-info/SOURCES.txt +1 -0
- {aither_kvcache-2.4.1 → aither_kvcache-2.4.2}/pyproject.toml +1 -1
- {aither_kvcache-2.4.1 → aither_kvcache-2.4.2}/tests/test_vllm_plugin.py +9 -1
- {aither_kvcache-2.4.1 → aither_kvcache-2.4.2}/LICENSE +0 -0
- {aither_kvcache-2.4.1 → aither_kvcache-2.4.2}/aither_kvcache/block_selector.py +0 -0
- {aither_kvcache-2.4.1 → aither_kvcache-2.4.2}/aither_kvcache/eviction_advisor.py +0 -0
- {aither_kvcache-2.4.1 → aither_kvcache-2.4.2}/aither_kvcache/fused_kv_update.py +0 -0
- {aither_kvcache-2.4.1 → aither_kvcache-2.4.2}/aither_kvcache/hybrid_quantizer.py +0 -0
- {aither_kvcache-2.4.1 → aither_kvcache-2.4.2}/aither_kvcache/kvcache_graph.py +0 -0
- {aither_kvcache-2.4.1 → aither_kvcache-2.4.2}/aither_kvcache/kvtransfer/__init__.py +0 -0
- {aither_kvcache-2.4.1 → aither_kvcache-2.4.2}/aither_kvcache/kvtransfer/mapper.py +0 -0
- {aither_kvcache-2.4.1 → aither_kvcache-2.4.2}/aither_kvcache/kvtransfer/rope.py +0 -0
- {aither_kvcache-2.4.1 → aither_kvcache-2.4.2}/aither_kvcache/persistence.py +0 -0
- {aither_kvcache-2.4.1 → aither_kvcache-2.4.2}/aither_kvcache/triattention/__init__.py +0 -0
- {aither_kvcache-2.4.1 → aither_kvcache-2.4.2}/aither_kvcache/triattention/attention.py +0 -0
- {aither_kvcache-2.4.1 → aither_kvcache-2.4.2}/aither_kvcache/triattention/cache.py +0 -0
- {aither_kvcache-2.4.1 → aither_kvcache-2.4.2}/aither_kvcache/triattention/calibration.py +0 -0
- {aither_kvcache-2.4.1 → aither_kvcache-2.4.2}/aither_kvcache/triattention/config.py +0 -0
- {aither_kvcache-2.4.1 → aither_kvcache-2.4.2}/aither_kvcache/triattention/encoder.py +0 -0
- {aither_kvcache-2.4.1 → aither_kvcache-2.4.2}/aither_kvcache/triattention/scorer.py +0 -0
- {aither_kvcache-2.4.1 → aither_kvcache-2.4.2}/aither_kvcache/triattention/spectral.py +0 -0
- {aither_kvcache-2.4.1 → aither_kvcache-2.4.2}/aither_kvcache/vllm/__init__.py +0 -0
- {aither_kvcache-2.4.1 → aither_kvcache-2.4.2}/aither_kvcache/vllm/backend.py +0 -0
- {aither_kvcache-2.4.1 → aither_kvcache-2.4.2}/aither_kvcache/vllm/cache.py +0 -0
- {aither_kvcache-2.4.1 → aither_kvcache-2.4.2}/aither_kvcache/vllm/engine.py +0 -0
- {aither_kvcache-2.4.1 → aither_kvcache-2.4.2}/aither_kvcache/vllm/eviction_plugin.py +0 -0
- {aither_kvcache-2.4.1 → aither_kvcache-2.4.2}/aither_kvcache/vllm/hooks.py +0 -0
- {aither_kvcache-2.4.1 → aither_kvcache-2.4.2}/aither_kvcache/vllm/isolation_status_endpoint.py +0 -0
- {aither_kvcache-2.4.1 → aither_kvcache-2.4.2}/aither_kvcache/vllm/kv_warmth_client.py +0 -0
- {aither_kvcache-2.4.1 → aither_kvcache-2.4.2}/aither_kvcache/vllm/plugin.py +0 -0
- {aither_kvcache-2.4.1 → aither_kvcache-2.4.2}/aither_kvcache/vllm/sitecustomize.py +0 -0
- {aither_kvcache-2.4.1 → aither_kvcache-2.4.2}/aither_kvcache/vllm/strata_shadow_plugin.py +0 -0
- {aither_kvcache-2.4.1 → aither_kvcache-2.4.2}/aither_kvcache/vllm/tenant_isolation.py +0 -0
- {aither_kvcache-2.4.1 → aither_kvcache-2.4.2}/aither_kvcache.egg-info/dependency_links.txt +0 -0
- {aither_kvcache-2.4.1 → aither_kvcache-2.4.2}/aither_kvcache.egg-info/entry_points.txt +0 -0
- {aither_kvcache-2.4.1 → aither_kvcache-2.4.2}/aither_kvcache.egg-info/requires.txt +0 -0
- {aither_kvcache-2.4.1 → aither_kvcache-2.4.2}/aither_kvcache.egg-info/top_level.txt +0 -0
- {aither_kvcache-2.4.1 → aither_kvcache-2.4.2}/setup.cfg +0 -0
- {aither_kvcache-2.4.1 → aither_kvcache-2.4.2}/tests/test_block_selector.py +0 -0
- {aither_kvcache-2.4.1 → aither_kvcache-2.4.2}/tests/test_core.py +0 -0
- {aither_kvcache-2.4.1 → aither_kvcache-2.4.2}/tests/test_fused.py +0 -0
- {aither_kvcache-2.4.1 → aither_kvcache-2.4.2}/tests/test_hybrid_hooks.py +0 -0
- {aither_kvcache-2.4.1 → aither_kvcache-2.4.2}/tests/test_kvcache_graph.py +0 -0
- {aither_kvcache-2.4.1 → aither_kvcache-2.4.2}/tests/test_persistence.py +0 -0
- {aither_kvcache-2.4.1 → aither_kvcache-2.4.2}/tests/test_triattention.py +0 -0
- {aither_kvcache-2.4.1 → aither_kvcache-2.4.2}/turboquant/__init__.py +0 -0
- {aither_kvcache-2.4.1 → aither_kvcache-2.4.2}/turboquant/bench.py +0 -0
- {aither_kvcache-2.4.1 → aither_kvcache-2.4.2}/turboquant/codebook.py +0 -0
- {aither_kvcache-2.4.1 → aither_kvcache-2.4.2}/turboquant/fused_attention.py +0 -0
- {aither_kvcache-2.4.1 → aither_kvcache-2.4.2}/turboquant/gb10_fp8_rescue.py +0 -0
- {aither_kvcache-2.4.1 → aither_kvcache-2.4.2}/turboquant/packing.py +0 -0
- {aither_kvcache-2.4.1 → aither_kvcache-2.4.2}/turboquant/quantizer.py +0 -0
- {aither_kvcache-2.4.1 → aither_kvcache-2.4.2}/turboquant/rotation.py +0 -0
- {aither_kvcache-2.4.1 → aither_kvcache-2.4.2}/turboquant/triton_ops.py +0 -0
|
@@ -1,6 +1,6 @@
|
|
|
1
1
|
Metadata-Version: 2.4
|
|
2
2
|
Name: aither-kvcache
|
|
3
|
-
Version: 2.4.
|
|
3
|
+
Version: 2.4.2
|
|
4
4
|
Summary: KV cache tooling for LLM inference — TurboQuant (vector quantization), TriAttention (spectral compression), and cross-model KV transfer (skip prefill by converting one model's cache into another's)
|
|
5
5
|
Author: Aitherium
|
|
6
6
|
License-Expression: CC-BY-4.0
|
|
@@ -44,7 +44,7 @@ Dynamic: license-file
|
|
|
44
44
|
|
|
45
45
|
**[Docs](https://aitherium.github.io/aitherkvcache/)** · [Source](https://github.com/Aitherium/aitherkvcache) · `pip install aither-kvcache` · [The Aither World](https://aitherium.github.io/)
|
|
46
46
|
|
|
47
|
-
> **The Aither World** is an operating system for agents — a Linux you can hand to one, the runtimes it works in, and the tools it works with. [awnix](https://github.com/Aitherium/awnix) is the Linux underneath it; **aitherkvcache** is one of its
|
|
47
|
+
> **The Aither World** is an operating system for agents — a Linux you can hand to one, the runtimes it works in, and the tools it works with. [awnix](https://github.com/Aitherium/awnix) is the Linux underneath it; **aitherkvcache** is one of its 69 bricks — each installs on its own, runs offline, and needs no account. 7 more are in development and 15 are planned.
|
|
48
48
|
>
|
|
49
49
|
> **Start here:** Quantize the cache on one model you already serve and measure the headroom.
|
|
50
50
|
|
|
@@ -726,9 +726,12 @@ Each installs on its own, works offline, and needs no account.
|
|
|
726
726
|
| [awgit](https://github.com/Aitherium/awgit) | that no one else is editing this file | a lease, refused at commit time if you do not hold it |
|
|
727
727
|
| [awdelphi](https://github.com/Aitherium/awdelphi) | one agent's confident take on a decision | the round trace, the anonymity, and who dissents |
|
|
728
728
|
| [awclassify](https://github.com/Aitherium/awclassify) | a filename, a folder, or whoever last touched it | doc_type, visibility, audience and topics, with the evidence lines that decided each |
|
|
729
|
+
| [awdecide](https://github.com/Aitherium/awdecide) | a hosted classifier's probability that never learns whether it was right | the decision, its probability, and the calibration curve from your own resolved outcomes |
|
|
729
730
|
| [awtoll](https://github.com/Aitherium/awtoll) | that your tooling is saving you context | the measured token cost of each tool call, and what the alternative cost |
|
|
730
731
|
| [awseal](https://github.com/Aitherium/awseal) | that the artifact came from who you think | an Ed25519 seal — the key that verifies is not the key that forges |
|
|
731
732
|
| [awshare](https://github.com/Aitherium/awshare) | that the download is intact | content-addressed bundles, verified on fetch |
|
|
733
|
+
| [awsuite](https://github.com/Aitherium/awsuite) | that an agent holding your mailbox will not send on its own | every send, draft, upload and create returns a dry-run until confirm is true |
|
|
734
|
+
| [awpermit](https://github.com/Aitherium/awpermit) | that an admitted agent will only do what its owner meant | an Ed25519 permit naming the agent, the principal, the door, the actions and the cap |
|
|
732
735
|
| [awnest](https://github.com/Aitherium/awnest) | that there is a person on the other end | a verdict with evidence, where "we could not tell" is not "yes" |
|
|
733
736
|
| [awrena](https://github.com/Aitherium/awrena) | a leaderboard someone can edit, and votes nobody counted | a scored duel with both answers kept, and a result bound to them |
|
|
734
737
|
| [awnboard](https://github.com/Aitherium/awnboard) | a share link anyone who sees it can use | an invitation addressed to one person, for one gate, revocable |
|
|
@@ -738,6 +741,7 @@ Each installs on its own, works offline, and needs no account.
|
|
|
738
741
|
| [awrelay](https://github.com/Aitherium/awrelay) | a SaaS in the middle of your agents | findings, alerts and coordination over your own transport |
|
|
739
742
|
| [awask](https://github.com/Aitherium/awask) | that anyone read the paragraph where you asked | the ask itself, with a button that steers the session that raised it |
|
|
740
743
|
| [awmail](https://github.com/Aitherium/awmail) | a mailbox somebody else can read | mail your agents send and receive over your own server |
|
|
744
|
+
| [awdiscord](https://github.com/Aitherium/awdiscord) | a self-bot on someone's account, or a bot that acts without asking | slash commands, rate limits, and approval buttons only your operators can press |
|
|
741
745
|
| [awswarm](https://github.com/Aitherium/awswarm) | that a model either fits your GPU or it doesn't run at all | a placement plan and an acquisition-probability estimate before you spend on a run |
|
|
742
746
|
| [awfind](https://github.com/Aitherium/awfind) | one vendor's idea of the web | results from whichever providers you configured |
|
|
743
747
|
| [awbrowse](https://github.com/Aitherium/awbrowse) | that the page said what you were told | the render, the DOM and the requests it made |
|
|
@@ -789,14 +793,17 @@ Every repository here is public. Each publishes an `aither-manifest.json` beside
|
|
|
789
793
|
| [awgraph](https://github.com/Aitherium/awgraph) | A semantic code graph for agents — AST + tree-sitter, call graphs | [docs](https://aitherium.github.io/awgraph/) |
|
|
790
794
|
| [awgit](https://github.com/Aitherium/awgit) | Semantic version control on top of git — edit-ops and leases | [docs](https://aitherium.github.io/awgit/) |
|
|
791
795
|
| [awdelphi](https://github.com/Aitherium/awdelphi) | Anonymous multi-round expert panels — a converged answer with a trace | [docs](https://aitherium.github.io/awdelphi/) |
|
|
792
|
-
| [awclassify](https://github.com/Aitherium/awclassify) | Classify any document -- what it is, who may read it, who it is for, what it is about |
|
|
796
|
+
| [awclassify](https://github.com/Aitherium/awclassify) | Classify any document -- what it is, who may read it, who it is for, what it is about | [docs](https://aitherium.github.io/awclassify/) |
|
|
797
|
+
| [awdecide](https://github.com/Aitherium/awdecide) | One typed-decision contract -- choice / score / bool with a probability -- over a ladder of backends you already run (rules, tiny local models, an LLM's logprobs), fail-closed, with a Brier ledger that resolves every decision against its outcome | [docs](https://aitherium.github.io/awdecide/) |
|
|
793
798
|
| [awtoll](https://github.com/Aitherium/awtoll) | What every tool call costs you in context, measured from your own transcripts | [docs](https://aitherium.github.io/awtoll/) |
|
|
794
799
|
| [awseal](https://github.com/Aitherium/awseal) | Sign an artifact so a stranger can verify it | [docs](https://aitherium.github.io/awseal/) |
|
|
795
800
|
| [awshare](https://github.com/Aitherium/awshare) | Publish an artifact and fetch it back verified | [docs](https://aitherium.github.io/awshare/) |
|
|
801
|
+
| [awsuite](https://github.com/Aitherium/awsuite) | Your Google Workspace as agent tools, and no write happens without a yes | [docs](https://aitherium.github.io/awsuite/) |
|
|
796
802
|
| [awdit](https://github.com/Aitherium/awdit) | An append-only audit trail whose gaps are DETECTABLE | [docs](https://aitherium.github.io/awdit/) |
|
|
797
803
|
| [awbac](https://github.com/Aitherium/awbac) | Role-based access control that fails closed and explains itself | [docs](https://aitherium.github.io/awbac/) |
|
|
798
804
|
| [awiam](https://github.com/Aitherium/awiam) | Who is this caller? A directory and session store that fails honestly | [docs](https://aitherium.github.io/awiam/) |
|
|
799
805
|
| [awtunnel](https://github.com/Aitherium/awtunnel) | Reach a service that has no public address | [docs](https://aitherium.github.io/awtunnel/) |
|
|
806
|
+
| [awpermit](https://github.com/Aitherium/awpermit) | Let an agent through your door with a scope, a spend cap, and someone accountable | [docs](https://aitherium.github.io/awpermit/) |
|
|
800
807
|
| [awnest](https://github.com/Aitherium/awnest) | Prove there is a human before you let them into the nest | [docs](https://aitherium.github.io/awnest/) |
|
|
801
808
|
| [awrena](https://github.com/Aitherium/awrena) | Put two agents head to head and get a verdict you can check | [docs](https://aitherium.github.io/awrena/) |
|
|
802
809
|
| [awnboard](https://github.com/Aitherium/awnboard) | A front gate you can put in front of anything, and hand someone the key to | [docs](https://aitherium.github.io/awnboard/) |
|
|
@@ -806,18 +813,19 @@ Every repository here is public. Each publishes an `aither-manifest.json` beside
|
|
|
806
813
|
| [awrelay](https://github.com/Aitherium/awrelay) | Portable agent messaging — findings, alerts, coordination | [docs](https://aitherium.github.io/awrelay/) |
|
|
807
814
|
| [awask](https://github.com/Aitherium/awask) | Your agent asks you a question — and acts on your answer | [docs](https://aitherium.github.io/awask/) |
|
|
808
815
|
| [awmail](https://github.com/Aitherium/awmail) | Give an agent an email address — send, and actually receive | [docs](https://aitherium.github.io/awmail/) |
|
|
816
|
+
| [awdiscord](https://github.com/Aitherium/awdiscord) | Put any agent on Discord as a real bot, with an operator in the loop | [docs](https://aitherium.github.io/awdiscord/) |
|
|
809
817
|
| [awnet](https://github.com/Aitherium/awnet) | The agentic web — agents host a mesh, and agents join one | [docs](https://aitherium.github.io/awnet/) |
|
|
810
|
-
| [awswarm](https://github.com/Aitherium/awswarm) | Run one model too big for any single GPU across a pool of small ones |
|
|
818
|
+
| [awswarm](https://github.com/Aitherium/awswarm) | Run one model too big for any single GPU across a pool of small ones | [docs](https://aitherium.github.io/awswarm/) |
|
|
811
819
|
| [awfind](https://github.com/Aitherium/awfind) | A portable search client — query, results, ranking | [docs](https://aitherium.github.io/awfind/) |
|
|
812
820
|
| [awbrowse](https://github.com/Aitherium/awbrowse) | A portable browser client — navigate, console, network, DOM, screenshot | [docs](https://aitherium.github.io/awbrowse/) |
|
|
813
821
|
| [awvoice](https://github.com/Aitherium/awvoice) | Hear and speak — transcribe audio, synthesize a voice | [docs](https://aitherium.github.io/awvoice/) |
|
|
814
822
|
| [awvision](https://github.com/Aitherium/awvision) | See an image — describe it, ask it a question, compare two | [docs](https://aitherium.github.io/awvision/) |
|
|
815
823
|
| [awscreen](https://github.com/Aitherium/awscreen) | See this machine — what is on screen, and where to click it | [docs](https://aitherium.github.io/awscreen/) |
|
|
816
|
-
| [awkit](https://github.com/Aitherium/awkit) | Render an agent panel from a tool result — one component, any React app |
|
|
817
|
-
| [awbeads](https://github.com/Aitherium/awbeads) | A spatial canvas for a page — arrange things, connect them, and keep the arrangement |
|
|
818
|
-
| [awbonsai](https://github.com/Aitherium/awbonsai) | Run a real model in the visitor's own browser — no server round trip, no upload |
|
|
824
|
+
| [awkit](https://github.com/Aitherium/awkit) | Render an agent panel from a tool result — one component, any React app | [docs](https://aitherium.github.io/awkit/) |
|
|
825
|
+
| [awbeads](https://github.com/Aitherium/awbeads) | A spatial canvas for a page — arrange things, connect them, and keep the arrangement | [docs](https://aitherium.github.io/awbeads/) |
|
|
826
|
+
| [awbonsai](https://github.com/Aitherium/awbonsai) | Run a real model in the visitor's own browser — no server round trip, no upload | [docs](https://aitherium.github.io/awbonsai/) |
|
|
819
827
|
| [awknowledge](https://github.com/Aitherium/awknowledge) | How to run a coding agent so the result survives — the laws, with evidence | [docs](https://aitherium.github.io/awknowledge/) |
|
|
820
|
-
| [awbrain](https://github.com/Aitherium/awbrain) | Your history as a wiki of linked markdown — claims pinned to the evidence |
|
|
828
|
+
| [awbrain](https://github.com/Aitherium/awbrain) | Your history as a wiki of linked markdown — claims pinned to the evidence | [docs](https://aitherium.github.io/awbrain/) |
|
|
821
829
|
| [gawbbonet](https://github.com/Aitherium/gawbbonet) | GobboNet campaigns with a real agent brain — scoped memory, graph recall | [docs](https://aitherium.github.io/gawbbonet/) |
|
|
822
830
|
| **aitherkvcache** _(you are here)_ | Near-optimal KV cache quantization for LLM inference — sub-byte compression | [docs](https://aitherium.github.io/aitherkvcache/) |
|
|
823
831
|
| [awrtifact](https://github.com/Aitherium/awrtifact) | Deliberately chunk artifacts into GitHub release assets — the productized aitherkvcache mirror lane | [docs](https://aitherium.github.io/awrtifact/) |
|
|
@@ -827,24 +835,26 @@ Every repository here is public. Each publishes an `aither-manifest.json` beside
|
|
|
827
835
|
| [awrecurse](https://github.com/Aitherium/awrecurse) | Answer a question over a context far larger than the window — recursively, with the trace kept | [docs](https://aitherium.github.io/awrecurse/) |
|
|
828
836
|
| [awprism](https://github.com/Aitherium/awprism) | Turn a failure into ranked hypotheses — and say what would confirm each one | [docs](https://aitherium.github.io/awprism/) |
|
|
829
837
|
| [awrepl](https://github.com/Aitherium/awrepl) | A REPL an agent can actually use — state that survives between turns | [docs](https://aitherium.github.io/awrepl/) |
|
|
830
|
-
| [awreport](https://github.com/Aitherium/awreport) | File a bug report that has already scrubbed your secrets and collapsed the duplicate |
|
|
838
|
+
| [awreport](https://github.com/Aitherium/awreport) | File a bug report that has already scrubbed your secrets and collapsed the duplicate | [docs](https://aitherium.github.io/awreport/) |
|
|
831
839
|
| [awresearch](https://github.com/Aitherium/awresearch) | Ask a research question, get a cited report you can check | [docs](https://aitherium.github.io/awresearch/) |
|
|
832
840
|
| [awfocus](https://github.com/Aitherium/awfocus) | See, search and steer every Claude session from one command | [docs](https://aitherium.github.io/awfocus/) |
|
|
833
841
|
| [awgym](https://github.com/Aitherium/awgym) | An ARC training gym — a game a world model can watch, and six roles that play through it | [docs](https://aitherium.github.io/awgym/) |
|
|
834
842
|
| [awpredict](https://github.com/Aitherium/awpredict) | Predict what your environment does next, and how surprised you were | [docs](https://aitherium.github.io/awpredict/) |
|
|
835
|
-
| [awevolve](https://github.com/Aitherium/awevolve) | Point an agent at a file and a command that scores it, and let it improve |
|
|
843
|
+
| [awevolve](https://github.com/Aitherium/awevolve) | Point an agent at a file and a command that scores it, and let it improve | [docs](https://aitherium.github.io/awevolve/) |
|
|
836
844
|
| [awsh](https://github.com/Aitherium/awsh) | Your terminal answers you -- type a question where a command would go | [docs](https://aitherium.github.io/awsh/) |
|
|
837
|
-
| [awmine](https://github.com/Aitherium/awmine) | Mine what your agents did -- outcomes, lessons and procedures out of the transcripts they left behind |
|
|
845
|
+
| [awmine](https://github.com/Aitherium/awmine) | Mine what your agents did -- outcomes, lessons and procedures out of the transcripts they left behind | [docs](https://aitherium.github.io/awmine/) |
|
|
838
846
|
| [awrise](https://github.com/Aitherium/awrise) | Wake an agent on a schedule, let it do one thing, and put it back to sleep | [docs](https://aitherium.github.io/awrise/) |
|
|
839
847
|
| [awkno](https://github.com/Aitherium/awkno) | The man page for the Aither World — every brick, stack and law, offline | [docs](https://aitherium.github.io/awkno/) |
|
|
840
848
|
| [awwall](https://github.com/Aitherium/awwall) | Say what a workload may reach, and watch everything else fail closed | [docs](https://aitherium.github.io/awwall/) |
|
|
841
|
-
| [awrouter](https://github.com/Aitherium/awrouter) | OpenRouter for your own fleet: pick a model backend by cost/latency/ capability, fail over, fit the context window, stream. Standalone, OpenAI-compatible, no Aither-specifics required to be valuable |
|
|
849
|
+
| [awrouter](https://github.com/Aitherium/awrouter) | OpenRouter for your own fleet: pick a model backend by cost/latency/ capability, fail over, fit the context window, stream. Standalone, OpenAI-compatible, no Aither-specifics required to be valuable | [docs](https://aitherium.github.io/awrouter/) |
|
|
842
850
|
| [awembed](https://github.com/Aitherium/awembed) | Train an embedding model that knows your corpus, and prove it beats the big one | [docs](https://aitherium.github.io/awembed/) |
|
|
843
851
|
| [awtax](https://github.com/Aitherium/awtax) | Turn any tax PDF -- returns, W-2, 1099, statements, even scans -- into structured data you can check | [docs](https://aitherium.github.io/awtax/) |
|
|
844
852
|
| [awflow](https://github.com/Aitherium/awflow) | A deterministic workflow runtime — chain agent calls with journal replay and budget control | [docs](https://aitherium.github.io/awflow/) |
|
|
845
853
|
| [awsettings](https://github.com/Aitherium/awsettings) | Your agent's permissions and config, following you to the next machine | [docs](https://aitherium.github.io/awsettings/) |
|
|
846
854
|
| [awavatar](https://github.com/Aitherium/awavatar) | One character spec in, a rigged, animated, multi-style avatar pack out | [docs](https://aitherium.github.io/awavatar/) |
|
|
847
855
|
|
|
856
|
+
**Built on** [REA (Reverse Engineer Anything)](https://github.com/morluto/rea) · [llama.cpp](https://github.com/ggml-org/llama.cpp) · [vLLM](https://github.com/vllm-project/vllm) · [ComfyUI](https://github.com/comfyanonymous/ComfyUI) · [CentOS Stream](https://www.centos.org/centos-stream/) · [Podman](https://github.com/containers/podman) · [Docker](https://github.com/moby/moby) · [LanceDB](https://github.com/lancedb/lancedb) · [WireGuard](https://www.wireguard.com/) · [FFmpeg](https://ffmpeg.org/) · [Blender + Rigify](https://www.blender.org/) · [headroom](https://github.com/headroomlabs-ai/headroom) · [SANA](https://github.com/NVlabs/Sana) · [Hunyuan3D](https://github.com/Tencent-Hunyuan/Hunyuan3D-2.1) · [repowise](https://github.com/repowise-dev/repowise) · [Playwright](https://github.com/microsoft/playwright) · [Chromium](https://www.chromium.org/) · [Next.js](https://github.com/vercel/next.js) · [React](https://github.com/facebook/react).
|
|
857
|
+
|
|
848
858
|
<div id="aither-constellation" data-self="aitherkvcache"></div>
|
|
849
859
|
<script src="aither-constellation.js"></script>
|
|
850
860
|
|
|
@@ -4,7 +4,7 @@
|
|
|
4
4
|
|
|
5
5
|
**[Docs](https://aitherium.github.io/aitherkvcache/)** · [Source](https://github.com/Aitherium/aitherkvcache) · `pip install aither-kvcache` · [The Aither World](https://aitherium.github.io/)
|
|
6
6
|
|
|
7
|
-
> **The Aither World** is an operating system for agents — a Linux you can hand to one, the runtimes it works in, and the tools it works with. [awnix](https://github.com/Aitherium/awnix) is the Linux underneath it; **aitherkvcache** is one of its
|
|
7
|
+
> **The Aither World** is an operating system for agents — a Linux you can hand to one, the runtimes it works in, and the tools it works with. [awnix](https://github.com/Aitherium/awnix) is the Linux underneath it; **aitherkvcache** is one of its 69 bricks — each installs on its own, runs offline, and needs no account. 7 more are in development and 15 are planned.
|
|
8
8
|
>
|
|
9
9
|
> **Start here:** Quantize the cache on one model you already serve and measure the headroom.
|
|
10
10
|
|
|
@@ -686,9 +686,12 @@ Each installs on its own, works offline, and needs no account.
|
|
|
686
686
|
| [awgit](https://github.com/Aitherium/awgit) | that no one else is editing this file | a lease, refused at commit time if you do not hold it |
|
|
687
687
|
| [awdelphi](https://github.com/Aitherium/awdelphi) | one agent's confident take on a decision | the round trace, the anonymity, and who dissents |
|
|
688
688
|
| [awclassify](https://github.com/Aitherium/awclassify) | a filename, a folder, or whoever last touched it | doc_type, visibility, audience and topics, with the evidence lines that decided each |
|
|
689
|
+
| [awdecide](https://github.com/Aitherium/awdecide) | a hosted classifier's probability that never learns whether it was right | the decision, its probability, and the calibration curve from your own resolved outcomes |
|
|
689
690
|
| [awtoll](https://github.com/Aitherium/awtoll) | that your tooling is saving you context | the measured token cost of each tool call, and what the alternative cost |
|
|
690
691
|
| [awseal](https://github.com/Aitherium/awseal) | that the artifact came from who you think | an Ed25519 seal — the key that verifies is not the key that forges |
|
|
691
692
|
| [awshare](https://github.com/Aitherium/awshare) | that the download is intact | content-addressed bundles, verified on fetch |
|
|
693
|
+
| [awsuite](https://github.com/Aitherium/awsuite) | that an agent holding your mailbox will not send on its own | every send, draft, upload and create returns a dry-run until confirm is true |
|
|
694
|
+
| [awpermit](https://github.com/Aitherium/awpermit) | that an admitted agent will only do what its owner meant | an Ed25519 permit naming the agent, the principal, the door, the actions and the cap |
|
|
692
695
|
| [awnest](https://github.com/Aitherium/awnest) | that there is a person on the other end | a verdict with evidence, where "we could not tell" is not "yes" |
|
|
693
696
|
| [awrena](https://github.com/Aitherium/awrena) | a leaderboard someone can edit, and votes nobody counted | a scored duel with both answers kept, and a result bound to them |
|
|
694
697
|
| [awnboard](https://github.com/Aitherium/awnboard) | a share link anyone who sees it can use | an invitation addressed to one person, for one gate, revocable |
|
|
@@ -698,6 +701,7 @@ Each installs on its own, works offline, and needs no account.
|
|
|
698
701
|
| [awrelay](https://github.com/Aitherium/awrelay) | a SaaS in the middle of your agents | findings, alerts and coordination over your own transport |
|
|
699
702
|
| [awask](https://github.com/Aitherium/awask) | that anyone read the paragraph where you asked | the ask itself, with a button that steers the session that raised it |
|
|
700
703
|
| [awmail](https://github.com/Aitherium/awmail) | a mailbox somebody else can read | mail your agents send and receive over your own server |
|
|
704
|
+
| [awdiscord](https://github.com/Aitherium/awdiscord) | a self-bot on someone's account, or a bot that acts without asking | slash commands, rate limits, and approval buttons only your operators can press |
|
|
701
705
|
| [awswarm](https://github.com/Aitherium/awswarm) | that a model either fits your GPU or it doesn't run at all | a placement plan and an acquisition-probability estimate before you spend on a run |
|
|
702
706
|
| [awfind](https://github.com/Aitherium/awfind) | one vendor's idea of the web | results from whichever providers you configured |
|
|
703
707
|
| [awbrowse](https://github.com/Aitherium/awbrowse) | that the page said what you were told | the render, the DOM and the requests it made |
|
|
@@ -749,14 +753,17 @@ Every repository here is public. Each publishes an `aither-manifest.json` beside
|
|
|
749
753
|
| [awgraph](https://github.com/Aitherium/awgraph) | A semantic code graph for agents — AST + tree-sitter, call graphs | [docs](https://aitherium.github.io/awgraph/) |
|
|
750
754
|
| [awgit](https://github.com/Aitherium/awgit) | Semantic version control on top of git — edit-ops and leases | [docs](https://aitherium.github.io/awgit/) |
|
|
751
755
|
| [awdelphi](https://github.com/Aitherium/awdelphi) | Anonymous multi-round expert panels — a converged answer with a trace | [docs](https://aitherium.github.io/awdelphi/) |
|
|
752
|
-
| [awclassify](https://github.com/Aitherium/awclassify) | Classify any document -- what it is, who may read it, who it is for, what it is about |
|
|
756
|
+
| [awclassify](https://github.com/Aitherium/awclassify) | Classify any document -- what it is, who may read it, who it is for, what it is about | [docs](https://aitherium.github.io/awclassify/) |
|
|
757
|
+
| [awdecide](https://github.com/Aitherium/awdecide) | One typed-decision contract -- choice / score / bool with a probability -- over a ladder of backends you already run (rules, tiny local models, an LLM's logprobs), fail-closed, with a Brier ledger that resolves every decision against its outcome | [docs](https://aitherium.github.io/awdecide/) |
|
|
753
758
|
| [awtoll](https://github.com/Aitherium/awtoll) | What every tool call costs you in context, measured from your own transcripts | [docs](https://aitherium.github.io/awtoll/) |
|
|
754
759
|
| [awseal](https://github.com/Aitherium/awseal) | Sign an artifact so a stranger can verify it | [docs](https://aitherium.github.io/awseal/) |
|
|
755
760
|
| [awshare](https://github.com/Aitherium/awshare) | Publish an artifact and fetch it back verified | [docs](https://aitherium.github.io/awshare/) |
|
|
761
|
+
| [awsuite](https://github.com/Aitherium/awsuite) | Your Google Workspace as agent tools, and no write happens without a yes | [docs](https://aitherium.github.io/awsuite/) |
|
|
756
762
|
| [awdit](https://github.com/Aitherium/awdit) | An append-only audit trail whose gaps are DETECTABLE | [docs](https://aitherium.github.io/awdit/) |
|
|
757
763
|
| [awbac](https://github.com/Aitherium/awbac) | Role-based access control that fails closed and explains itself | [docs](https://aitherium.github.io/awbac/) |
|
|
758
764
|
| [awiam](https://github.com/Aitherium/awiam) | Who is this caller? A directory and session store that fails honestly | [docs](https://aitherium.github.io/awiam/) |
|
|
759
765
|
| [awtunnel](https://github.com/Aitherium/awtunnel) | Reach a service that has no public address | [docs](https://aitherium.github.io/awtunnel/) |
|
|
766
|
+
| [awpermit](https://github.com/Aitherium/awpermit) | Let an agent through your door with a scope, a spend cap, and someone accountable | [docs](https://aitherium.github.io/awpermit/) |
|
|
760
767
|
| [awnest](https://github.com/Aitherium/awnest) | Prove there is a human before you let them into the nest | [docs](https://aitherium.github.io/awnest/) |
|
|
761
768
|
| [awrena](https://github.com/Aitherium/awrena) | Put two agents head to head and get a verdict you can check | [docs](https://aitherium.github.io/awrena/) |
|
|
762
769
|
| [awnboard](https://github.com/Aitherium/awnboard) | A front gate you can put in front of anything, and hand someone the key to | [docs](https://aitherium.github.io/awnboard/) |
|
|
@@ -766,18 +773,19 @@ Every repository here is public. Each publishes an `aither-manifest.json` beside
|
|
|
766
773
|
| [awrelay](https://github.com/Aitherium/awrelay) | Portable agent messaging — findings, alerts, coordination | [docs](https://aitherium.github.io/awrelay/) |
|
|
767
774
|
| [awask](https://github.com/Aitherium/awask) | Your agent asks you a question — and acts on your answer | [docs](https://aitherium.github.io/awask/) |
|
|
768
775
|
| [awmail](https://github.com/Aitherium/awmail) | Give an agent an email address — send, and actually receive | [docs](https://aitherium.github.io/awmail/) |
|
|
776
|
+
| [awdiscord](https://github.com/Aitherium/awdiscord) | Put any agent on Discord as a real bot, with an operator in the loop | [docs](https://aitherium.github.io/awdiscord/) |
|
|
769
777
|
| [awnet](https://github.com/Aitherium/awnet) | The agentic web — agents host a mesh, and agents join one | [docs](https://aitherium.github.io/awnet/) |
|
|
770
|
-
| [awswarm](https://github.com/Aitherium/awswarm) | Run one model too big for any single GPU across a pool of small ones |
|
|
778
|
+
| [awswarm](https://github.com/Aitherium/awswarm) | Run one model too big for any single GPU across a pool of small ones | [docs](https://aitherium.github.io/awswarm/) |
|
|
771
779
|
| [awfind](https://github.com/Aitherium/awfind) | A portable search client — query, results, ranking | [docs](https://aitherium.github.io/awfind/) |
|
|
772
780
|
| [awbrowse](https://github.com/Aitherium/awbrowse) | A portable browser client — navigate, console, network, DOM, screenshot | [docs](https://aitherium.github.io/awbrowse/) |
|
|
773
781
|
| [awvoice](https://github.com/Aitherium/awvoice) | Hear and speak — transcribe audio, synthesize a voice | [docs](https://aitherium.github.io/awvoice/) |
|
|
774
782
|
| [awvision](https://github.com/Aitherium/awvision) | See an image — describe it, ask it a question, compare two | [docs](https://aitherium.github.io/awvision/) |
|
|
775
783
|
| [awscreen](https://github.com/Aitherium/awscreen) | See this machine — what is on screen, and where to click it | [docs](https://aitherium.github.io/awscreen/) |
|
|
776
|
-
| [awkit](https://github.com/Aitherium/awkit) | Render an agent panel from a tool result — one component, any React app |
|
|
777
|
-
| [awbeads](https://github.com/Aitherium/awbeads) | A spatial canvas for a page — arrange things, connect them, and keep the arrangement |
|
|
778
|
-
| [awbonsai](https://github.com/Aitherium/awbonsai) | Run a real model in the visitor's own browser — no server round trip, no upload |
|
|
784
|
+
| [awkit](https://github.com/Aitherium/awkit) | Render an agent panel from a tool result — one component, any React app | [docs](https://aitherium.github.io/awkit/) |
|
|
785
|
+
| [awbeads](https://github.com/Aitherium/awbeads) | A spatial canvas for a page — arrange things, connect them, and keep the arrangement | [docs](https://aitherium.github.io/awbeads/) |
|
|
786
|
+
| [awbonsai](https://github.com/Aitherium/awbonsai) | Run a real model in the visitor's own browser — no server round trip, no upload | [docs](https://aitherium.github.io/awbonsai/) |
|
|
779
787
|
| [awknowledge](https://github.com/Aitherium/awknowledge) | How to run a coding agent so the result survives — the laws, with evidence | [docs](https://aitherium.github.io/awknowledge/) |
|
|
780
|
-
| [awbrain](https://github.com/Aitherium/awbrain) | Your history as a wiki of linked markdown — claims pinned to the evidence |
|
|
788
|
+
| [awbrain](https://github.com/Aitherium/awbrain) | Your history as a wiki of linked markdown — claims pinned to the evidence | [docs](https://aitherium.github.io/awbrain/) |
|
|
781
789
|
| [gawbbonet](https://github.com/Aitherium/gawbbonet) | GobboNet campaigns with a real agent brain — scoped memory, graph recall | [docs](https://aitherium.github.io/gawbbonet/) |
|
|
782
790
|
| **aitherkvcache** _(you are here)_ | Near-optimal KV cache quantization for LLM inference — sub-byte compression | [docs](https://aitherium.github.io/aitherkvcache/) |
|
|
783
791
|
| [awrtifact](https://github.com/Aitherium/awrtifact) | Deliberately chunk artifacts into GitHub release assets — the productized aitherkvcache mirror lane | [docs](https://aitherium.github.io/awrtifact/) |
|
|
@@ -787,24 +795,26 @@ Every repository here is public. Each publishes an `aither-manifest.json` beside
|
|
|
787
795
|
| [awrecurse](https://github.com/Aitherium/awrecurse) | Answer a question over a context far larger than the window — recursively, with the trace kept | [docs](https://aitherium.github.io/awrecurse/) |
|
|
788
796
|
| [awprism](https://github.com/Aitherium/awprism) | Turn a failure into ranked hypotheses — and say what would confirm each one | [docs](https://aitherium.github.io/awprism/) |
|
|
789
797
|
| [awrepl](https://github.com/Aitherium/awrepl) | A REPL an agent can actually use — state that survives between turns | [docs](https://aitherium.github.io/awrepl/) |
|
|
790
|
-
| [awreport](https://github.com/Aitherium/awreport) | File a bug report that has already scrubbed your secrets and collapsed the duplicate |
|
|
798
|
+
| [awreport](https://github.com/Aitherium/awreport) | File a bug report that has already scrubbed your secrets and collapsed the duplicate | [docs](https://aitherium.github.io/awreport/) |
|
|
791
799
|
| [awresearch](https://github.com/Aitherium/awresearch) | Ask a research question, get a cited report you can check | [docs](https://aitherium.github.io/awresearch/) |
|
|
792
800
|
| [awfocus](https://github.com/Aitherium/awfocus) | See, search and steer every Claude session from one command | [docs](https://aitherium.github.io/awfocus/) |
|
|
793
801
|
| [awgym](https://github.com/Aitherium/awgym) | An ARC training gym — a game a world model can watch, and six roles that play through it | [docs](https://aitherium.github.io/awgym/) |
|
|
794
802
|
| [awpredict](https://github.com/Aitherium/awpredict) | Predict what your environment does next, and how surprised you were | [docs](https://aitherium.github.io/awpredict/) |
|
|
795
|
-
| [awevolve](https://github.com/Aitherium/awevolve) | Point an agent at a file and a command that scores it, and let it improve |
|
|
803
|
+
| [awevolve](https://github.com/Aitherium/awevolve) | Point an agent at a file and a command that scores it, and let it improve | [docs](https://aitherium.github.io/awevolve/) |
|
|
796
804
|
| [awsh](https://github.com/Aitherium/awsh) | Your terminal answers you -- type a question where a command would go | [docs](https://aitherium.github.io/awsh/) |
|
|
797
|
-
| [awmine](https://github.com/Aitherium/awmine) | Mine what your agents did -- outcomes, lessons and procedures out of the transcripts they left behind |
|
|
805
|
+
| [awmine](https://github.com/Aitherium/awmine) | Mine what your agents did -- outcomes, lessons and procedures out of the transcripts they left behind | [docs](https://aitherium.github.io/awmine/) |
|
|
798
806
|
| [awrise](https://github.com/Aitherium/awrise) | Wake an agent on a schedule, let it do one thing, and put it back to sleep | [docs](https://aitherium.github.io/awrise/) |
|
|
799
807
|
| [awkno](https://github.com/Aitherium/awkno) | The man page for the Aither World — every brick, stack and law, offline | [docs](https://aitherium.github.io/awkno/) |
|
|
800
808
|
| [awwall](https://github.com/Aitherium/awwall) | Say what a workload may reach, and watch everything else fail closed | [docs](https://aitherium.github.io/awwall/) |
|
|
801
|
-
| [awrouter](https://github.com/Aitherium/awrouter) | OpenRouter for your own fleet: pick a model backend by cost/latency/ capability, fail over, fit the context window, stream. Standalone, OpenAI-compatible, no Aither-specifics required to be valuable |
|
|
809
|
+
| [awrouter](https://github.com/Aitherium/awrouter) | OpenRouter for your own fleet: pick a model backend by cost/latency/ capability, fail over, fit the context window, stream. Standalone, OpenAI-compatible, no Aither-specifics required to be valuable | [docs](https://aitherium.github.io/awrouter/) |
|
|
802
810
|
| [awembed](https://github.com/Aitherium/awembed) | Train an embedding model that knows your corpus, and prove it beats the big one | [docs](https://aitherium.github.io/awembed/) |
|
|
803
811
|
| [awtax](https://github.com/Aitherium/awtax) | Turn any tax PDF -- returns, W-2, 1099, statements, even scans -- into structured data you can check | [docs](https://aitherium.github.io/awtax/) |
|
|
804
812
|
| [awflow](https://github.com/Aitherium/awflow) | A deterministic workflow runtime — chain agent calls with journal replay and budget control | [docs](https://aitherium.github.io/awflow/) |
|
|
805
813
|
| [awsettings](https://github.com/Aitherium/awsettings) | Your agent's permissions and config, following you to the next machine | [docs](https://aitherium.github.io/awsettings/) |
|
|
806
814
|
| [awavatar](https://github.com/Aitherium/awavatar) | One character spec in, a rigged, animated, multi-style avatar pack out | [docs](https://aitherium.github.io/awavatar/) |
|
|
807
815
|
|
|
816
|
+
**Built on** [REA (Reverse Engineer Anything)](https://github.com/morluto/rea) · [llama.cpp](https://github.com/ggml-org/llama.cpp) · [vLLM](https://github.com/vllm-project/vllm) · [ComfyUI](https://github.com/comfyanonymous/ComfyUI) · [CentOS Stream](https://www.centos.org/centos-stream/) · [Podman](https://github.com/containers/podman) · [Docker](https://github.com/moby/moby) · [LanceDB](https://github.com/lancedb/lancedb) · [WireGuard](https://www.wireguard.com/) · [FFmpeg](https://ffmpeg.org/) · [Blender + Rigify](https://www.blender.org/) · [headroom](https://github.com/headroomlabs-ai/headroom) · [SANA](https://github.com/NVlabs/Sana) · [Hunyuan3D](https://github.com/Tencent-Hunyuan/Hunyuan3D-2.1) · [repowise](https://github.com/repowise-dev/repowise) · [Playwright](https://github.com/microsoft/playwright) · [Chromium](https://www.chromium.org/) · [Next.js](https://github.com/vercel/next.js) · [React](https://github.com/facebook/react).
|
|
817
|
+
|
|
808
818
|
<div id="aither-constellation" data-self="aitherkvcache"></div>
|
|
809
819
|
<script src="aither-constellation.js"></script>
|
|
810
820
|
|
|
@@ -56,4 +56,11 @@ __all__ = [
|
|
|
56
56
|
# TriAttention — lazy-loaded via subpackage
|
|
57
57
|
# from aither_kvcache.triattention import TriAttention, TriAttentionConfig
|
|
58
58
|
]
|
|
59
|
-
|
|
59
|
+
try:
|
|
60
|
+
# The installed distribution's version, so this can never again disagree with
|
|
61
|
+
# pyproject.toml (2.4.1 shipped reporting "2.1.0").
|
|
62
|
+
from importlib.metadata import PackageNotFoundError, version as _dist_version
|
|
63
|
+
|
|
64
|
+
__version__ = _dist_version("aither-kvcache")
|
|
65
|
+
except PackageNotFoundError: # running from a source checkout, not installed
|
|
66
|
+
__version__ = "0+unknown"
|