ciaacc-aether-token

Provider-neutral n8n nodes that reduce real AI Agent token usage with plug-and-play savings profiles.

Package Information

Released: 2 mo ago
Downloads: 0 weekly / 0 monthly
Latest Version: 0.1.0

Available Nodes

Documentation generating…

Aether Token Compress icon
Aether Token Compress
Reduce prompt, JSON, logs, HTML, history, and API payload tokens before AI reads them
Aether Token Model icon
Aether Token Model
Reduce the real prompt, memory, and tool schemas sent to any connected chat model
Aether Token Tools icon
Aether Token Tools
Reduce tool schemas, duplicate calls, and large tool results before the Agent reads them

Documentation

CIAACC Aether Token

Three plug-and-play n8n nodes that reduce the payload that reaches AI providers. The package is model-neutral and does not call a second compression model.

Install

In n8n, open Settings > Community Nodes, choose Install, and enter:

n8n-nodes-ciaacc-aether-token

For self-hosted command-line installation:

npm install n8n-nodes-ciaacc-aether-token

Nodes

Node Place it What it reduces
Aether Token Model Chat Model -> Aether Token Model -> AI Agent System prompts, old memory, inactive tools, repeated calls, and safe trivial turns
Aether Token Tools AI Tool -> Aether Token Tools -> AI Agent Repeated schemas, duplicate safe reads, and large tool results
Aether Token Compress Before an Agent, Chain, Tool, or model input JSON, prompts, history, logs, HTML, and API responses

The default configuration needs only one choice: Savings Power.

Savings Power

Profile Priority Eligible target
Hero Maximum fidelity 0-15%
Demigod Safe deterministic savings 15-30%
Olympian Balanced; recommended 30-50%
Titan High reversible savings 50-70%
Primordial Maximum eligible savings 70-85%

Targets are not guarantees. If the context contains no safely removable tokens, Aether returns the original. Highly repetitive or virtualized data can exceed the target.

Quality rules

  • Protect IDs, dates, values, URLs, emails, negations, tool protocol, and configured JSON paths.
  • Prefer lossless cleanup, reversible JSON packing, lazy tools, and exact cache reuse.
  • Store virtualized originals in a short-lived local resource and expose exact retrieval to Aether Token Model.
  • Reject negative optimization and fall back to the original when a protected fact is missing.
  • Report provider input/output usage when the connected provider returns usage metadata; otherwise label values as estimates.

Local proof

npm ci
npm run check
powershell -ExecutionPolicy Bypass -File scripts/start-local.ps1

Open http://localhost:5678 and run [LAB] Aether Token - Proof 6K. The workflow uses fictitious data and calls no external API or LLM.

The local runtime is stored outside the repository at %LOCALAPPDATA%\CIAACC-Aether-Token\n8n.

Benchmark

npm run benchmark

The benchmark measures the payload received by a local metered model. It does not claim provider billing savings unless provider usage metadata is available.

Current limits

  • Savings percentages are eligible targets, not guarantees.
  • Provider-billed savings require usage metadata from the connected model.
  • Semantic selection is enabled only in stronger profiles and falls back when protected facts fail validation.
  • Virtualized resources and exact-response cache are process-local in 0.1.0; queue-mode deployments need a shared store in a future release.
  • Test every profile against representative evals before production rollout.

Links

Discussion