mirror of
https://github.com/NousResearch/hermes-agent.git
synced 2026-11-06 05:17:15 +00:00
Patrick Suzuki
92f5cf5a7d
Repin to an asset-only child of cd370f7c6ef9279ae044269943ce2b7532a8c753 and use its immutable raw PNG URL. Keep version 0.1.0, all disclosures and capabilities, community tier, and existing screenshots unchanged.
27 lines
2.5 KiB
YAML
27 lines
2.5 KiB
YAML
name: token-lens
|
|
repo: https://github.com/patrickluvsoj/hermes-token-lens
|
|
sha: a6f20c2df26840bcfc11d8027c9fab6a847cf5dc
|
|
image: https://raw.githubusercontent.com/patrickluvsoj/hermes-token-lens/a6f20c2df26840bcfc11d8027c9fab6a847cf5dc/docs/banner.png
|
|
description: "Token consumption analytics for Hermes: records exact per-call usage from pre/post_api_request hooks, decomposes every request into categories (system prompt, built-in and per-MCP-server tool schemas, skills, memory, history, tool results, output, reasoning) calibrated to billed totals, and adds a Token Lens dashboard tab plus a Sessions-page card with deterministic and AI-evaluated token-reduction suggestions; `hermes token-lens refresh` runs a suggestion refresh from the CLI. Disclosure — records per-API-call token counts (no message text) from pre/post_api_request hooks into token_lens.db at the Hermes home root (per-call rows kept 90 days, rollups kept), and reads Hermes's state.db read-only for session token totals and, during the optional dashboard history import, message text for estimation only; after 10 recorded sessions it automatically calls your configured main model through Hermes's plugin LLM facade every 5 sessions (up to two calls per refresh, capped at 50,000 tokens by meta_budget_tokens), sending aggregated token statistics plus the names of your enabled MCP servers and plugins, and auto-applies the model's proposed category regex rules and rubric changes (logged to token_lens.EVOLUTION.md; evolve_rules: false turns this off); the dashboard's refresh button spawns a detached `hermes token-lens refresh` process that logs to logs/token-lens-refresh.log; its data paths follow the HERMES_HOME environment variable rather than the active profile, so in a multi-profile gateway every profile records into the launch profile's database; no third-party network calls, no telemetry."
|
|
maintainer: patrickluvsoj
|
|
tier: community
|
|
category: general
|
|
requires_hermes: ">=0.16.0"
|
|
docs_url: https://github.com/patrickluvsoj/hermes-token-lens#readme
|
|
version: "0.1.0"
|
|
screenshots:
|
|
- https://raw.githubusercontent.com/patrickluvsoj/hermes-token-lens/cd370f7c6ef9279ae044269943ce2b7532a8c753/docs/screenshots/token-lens-tab.png
|
|
- https://raw.githubusercontent.com/patrickluvsoj/hermes-token-lens/cd370f7c6ef9279ae044269943ce2b7532a8c753/docs/screenshots/entry-card.png
|
|
readme: true
|
|
platforms: []
|
|
capabilities:
|
|
provides_tools: []
|
|
provides_hooks:
|
|
- on_session_finalize
|
|
- on_session_start
|
|
- post_api_request
|
|
- pre_api_request
|
|
- pre_llm_call
|
|
provides_middleware: []
|
|
requires_env: []
|