by2kb 0.3.0__tar.gz

by2kb-0.3.0/.env.example

@@ -0,0 +1,73 @@
+# by2kb secrets — copy this file to `.env` (gitignored) and fill in real values.
+# Never commit the filled-in `.env`. Process environment variables take
+# precedence over entries in `.env`.
+#
+# `.env` is auto-loaded from (first match wins):
+#   1. $BY2KB_ENV_FILE
+#   2. <BY2KB_HOME>/.env        (default home: ~/.by2kb)
+#   3. ./.env                   (current working directory)
+
+# ---------------------------------------------------------------------------
+# ASR provider: doubao_auc  (Bilibili audio -> text)
+# Two independent credential domains, even if they belong to one Volcengine
+# account. See docs/reference/doubao-auc-tos-asr.md for the full flow.
+# ---------------------------------------------------------------------------
+
+# Volcengine TOS (private staging bucket for audio before ASR)
+VOLC_ACCESS_KEY_ID=
+VOLC_SECRET_ACCESS_KEY=
+TOS_BUCKET=
+# Optional, defaults shown:
+# TOS_REGION=ap-southeast-1
+# TOS_S3_ENDPOINT=tos-s3-ap-southeast-1.volces.com
+
+# Doubao AUC async file transcription (openspeech.bytedance.com)
+# Preferred new-console authentication:
+DOUBAO_API_KEY=
+# Optional legacy-console fallback:
+DOUBAO_APPID=
+DOUBAO_ACCESS_TOKEN=
+# Optional; verify against the speech product actually purchased:
+# DOUBAO_RESOURCE_ID=volc.seedasr.auc
+
+# ---------------------------------------------------------------------------
+# Local ASR provider: faster_whisper
+# Install the optional dependency and model explicitly before selecting it:
+#   pipx inject by2kb "faster-whisper>=1.2.1,<2"
+#   by2kb models install large-v3-turbo
+# Configure [asr] provider = "faster_whisper" in config.toml.
+# ---------------------------------------------------------------------------
+# BY2KB_WHISPER_MODEL=large-v3-turbo
+# BY2KB_WHISPER_DEVICE=auto
+# BY2KB_WHISPER_COMPUTE_TYPE=default
+# BY2KB_WHISPER_MODEL_DIR=~/.by2kb/models/faster-whisper
+# BY2KB_WHISPER_VAD_FILTER=true
+# BY2KB_WHISPER_BEAM_SIZE=5
+# BY2KB_WHISPER_CPU_THREADS=0
+
+# ---------------------------------------------------------------------------
+# LLM for summary enrichment (produces short.<title>.md + long.<title>.md).
+# Optional: without it the pipeline still completes with raw.<title>.md only.
+# This is API-key authentication; no browser or consumer-account login is used.
+# ---------------------------------------------------------------------------
+BY2KB_LLM_API_KEY=
+BY2KB_LLM_MODEL=
+# Optional, defaults to the Volcengine Ark preset:
+# BY2KB_LLM_BASE_URL=https://ark.cn-beijing.volces.com/api/v3
+# OpenAI-compatible alternative:
+# BY2KB_LLM_BASE_URL=https://api.openai.com/v1
+
+# Optional skill overrides. Packaged defaults are used when omitted.
+# BY2KB_ABSTRACT_SKILL=short-video-abstract
+# BY2KB_STUDY_SKILL=default-video-digest
+
+# ---------------------------------------------------------------------------
+# Storage locations (all optional)
+# ---------------------------------------------------------------------------
+# BY2KB_HOME=~/.by2kb
+# BY2KB_LIBRARY_ROOT=<where raw/abstract/updated artifacts are published>
+# BY2KB_DB_PATH=<job store sqlite path>
+
+# Optional comma-separated URL source-provider order. The default is the native
+# Bilibili provider only. Configure yt_dlp in config.toml for its detailed policy.
+# BY2KB_SOURCE_PROVIDERS=bilibili_native,yt_dlp

by2kb-0.3.0/.gitignore

@@ -0,0 +1,29 @@
+# Secrets and local configuration
+.env
+.env.*
+!.env.example
+*.pem
+*.key
+
+# Runtime data
+library/
+data/
+cache/
+*.db
+*.sqlite*
+
+# Dependencies and build output
+node_modules/
+.venv/
+__pycache__/
+dist/
+build/
+coverage/
+.pytest_cache/
+.test-tmp-*/
+*.egg-info/
+
+# Editors and OS
+.DS_Store
+.idea/
+.vscode/

by2kb-0.3.0/CHANGELOG.md

@@ -0,0 +1,49 @@
+# Changelog
+
+## Unreleased
+
+## 0.3.0 - 2026-08-29
+
+- Add a configurable source-provider registry, adapt native Bilibili ingestion to the
+  shared contract, and add an optional yt-dlp transcript/audio provider.
+- Complete the YouTube URL journey with caption-first ingestion, configured ASR
+  fallback, stable identity/provenance, and Hermes interception.
+- Add a pluggable ASR provider registry with deterministic explicit selection and
+  ordered `auto` fallback.
+- Add optional local `faster-whisper` transcription with timestamped segments,
+  configurable runtime settings, and explicit model status/install commands.
+- Accept local audio and video paths in `by2kb ingest`, with content-addressed
+  identity, ffmpeg audio extraction, and privacy-safe filename provenance.
+- Expand guided initialization for local/cloud ASR and add read-only `by2kb doctor`
+  diagnostics with a stable Agent-readable JSON report.
+- Add deterministic transcript-quality metrics and gates so unusable transcripts stop
+  before enrichment while borderline outputs carry a visible warning.
+- Add a cached long-form enrichment pipeline that plans segment-safe chunks,
+  recursively reduces grounded notes, and records a public hierarchy trace.
+- Add a staged Agent enrichment provider using bounded `next`/`submit` operations,
+  allowing Hermes subscription-authenticated runtimes to execute the shared plan.
+- Add a versioned Agent task-control protocol with stable status snapshots, bounded
+  waits, cooperative cancellation, and checkpoint-aware retry.
+- Persist the selected ASR provider, model, and runtime provenance in source and
+  transcript artifacts.
+
+## 0.2.1 - 2026-08-25
+
+- Name durable Markdown artifacts `raw.<title>.md`, `short.<title>.md`, and
+  `long.<title>.md`; safely retire fixed legacy names and unambiguous prior
+  suffix-based names on republish.
+- Fall back through Bilibili backup audio CDNs when the preferred CDN returns an
+  HTTP or transport failure.
+- Retry transient Doubao AUC failures per long-audio chunk and persist successful
+  chunk transcripts so a resumed ingestion does not resubmit completed work.
+- Classify queued, rate-limited, overloaded, and HTTP failure responses correctly;
+  harden concurrent checkpoint writes and cache invalidation.
+
+## 0.2.0 - 2026-08-23
+
+- Add agent-first enrichment with durable `claim`, `complete`, and `fail` commands.
+- Add short-abstract and long-form study-note execution through external agent hosts.
+- Add the Hermes reference plugin, authorized URL interception, and host-owned LLM use.
+- Add the bundled `video-to-knowledge` Hermes Skill and one-command plugin installer.
+- Add guided `by2kb init` configuration for TOS, Doubao ASR, LLM mode, and local KB.
+- Keep standalone API enrichment and transcript-only operation as supported executors.

by2kb-0.3.0/LICENSE

@@ -0,0 +1,21 @@
+MIT License
+
+Copyright (c) 2026 Charles
+
+Permission is hereby granted, free of charge, to any person obtaining a copy
+of this software and associated documentation files (the "Software"), to deal
+in the Software without restriction, including without limitation the rights
+to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
+copies of the Software, and to permit persons to whom the Software is
+furnished to do so, subject to the following conditions:
+
+The above copyright notice and this permission notice shall be included in all
+copies or substantial portions of the Software.
+
+THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
+AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
+OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+SOFTWARE.