.env.example · aaronrene/knowtation

1

# Copy to .env and set your keys. Do not commit .env.

2

# OPENAI_API_KEY=sk-your-key-here

3

#

4

# ── Two LLM lanes (different keys / providers) ─────────────────────────────

5

# 1) EMBEDDINGS = Meaning search + indexer (vectors). NOT Anthropic — use Ollama,

6

# OpenAI, Voyage, or DeepInfra. Set in config/local.yaml embedding: { provider, model } and/or:

7

# EMBEDDING_PROVIDER=ollama|openai|voyage|deepinfra EMBEDDING_MODEL=...

8

# Voyage (cloud, non-OpenAI): EMBEDDING_PROVIDER=voyage + VOYAGE_API_KEY

9

# DeepInfra (single-key with chat above; OpenAI-compatible API):

10

# EMBEDDING_PROVIDER=deepinfra + DEEPINFRA_API_KEY

11

# EMBEDDING_MODEL=BAAI/bge-large-en-v1.5 # default; 1024-dim

12

# Other options: Qwen/Qwen3-Embedding-8B (4096), intfloat/multilingual-e5-large-instruct (1024)

13

# NOTE: Switching embedding provider OR dimension requires a vault re-index.

14

# 2) CHAT = proposal Enrich / review hints / completeChat.

15

# Provider order (lib/llm-complete.mjs):

16

# a) KNOWTATION_CHAT_PROVIDER=deepinfra → DeepInfra (Qwen/Llama/Mistral) with

17

# OpenAI/Anthropic fallback if their keys are also set. Single OpenAI-compatible bill;

18

# same DEEPINFRA_API_KEY also works for OpenClaw orchestration (image gen, embeddings, TTS).

19

# b) KNOWTATION_CHAT_PROVIDER=openrouter → OpenRouter (BYO key, OpenAI-compatible).

20

# Requires OPENROUTER_API_KEY. NO fallback to a managed lane on failure: this is a

21

# "bring your own provider" lane (you pay OpenRouter directly; never metered against

22

# Knowtation packs), so an error surfaces instead of silently re-routing your note

23

# text to a metered provider. Explicit-only: adding OPENROUTER_API_KEY alone never

24

# changes the provider for an existing deploy. See

25

# docs/COMPANION-APP-MODEL-ROUTING-AND-ENRICHMENT-ARCHITECTURE.md §4/§6.

26

# c) KNOWTATION_CHAT_PROVIDER=openai|anthropic → lock to that provider (no fallback).

27

# d) Implicit DeepInfra: DEEPINFRA_API_KEY set AND no OpenAI/Anthropic key → DeepInfra.

28

# e) Default: OpenAI → Anthropic → Ollama (unchanged for existing deploys).

29

# KNOWTATION_CHAT_PREFER_ANTHROPIC=1 flips OpenAI/Anthropic order when both are set.

30

# ANTHROPIC_API_KEY=

31

# ANTHROPIC_CHAT_MODEL=claude-3-5-haiku-20241022

32

# KNOWTATION_CHAT_PREFER_ANTHROPIC=1

33

# DEEPINFRA_API_KEY=

34

# DEEPINFRA_CHAT_MODEL=Qwen/Qwen2.5-72B-Instruct # default; cheaper hint model: meta-llama/Meta-Llama-3.1-8B-Instruct

35

# KNOWTATION_CHAT_PROVIDER=deepinfra # explicit lock; omit for default OpenAI-first behavior

36

# OPENROUTER_API_KEY= # BYO key for KNOWTATION_CHAT_PROVIDER=openrouter

37

# OPENROUTER_CHAT_MODEL=openai/gpt-4o-mini # default; any OpenRouter model slug, e.g. anthropic/claude-3.5-haiku

38

# OPENROUTER_SITE_URL= # optional attribution → HTTP-Referer header

39

# OPENROUTER_APP_TITLE= # optional attribution → X-Title header

40

# VOYAGE_API_KEY= # when embedding.provider is voyage (see docs/setup.md)

41

# Hub (Phase 11): required for npm run hub (self-hosted). See docs/TWO-PATHS-HOSTED-AND-SELF-HOSTED.md#quick-start-self-hosted

42

# HUB_JWT_SECRET=your-long-random-secret

43

# KNOWTATION_VAULT_PATH=/absolute/path/to/vault (or set in config/local.yaml)

44

#

45

# Gateway → canister authentication (required for production).

46

# Set a strong random secret (64+ hex chars), then call admin_set_gateway_auth_secret on the canister.

47

# CANISTER_AUTH_SECRET=

48

#

49

# Full operator export (all users; hub canister must have secret set — see docs/OPERATOR-BACKUP.md)

50

# KNOWTATION_OPERATOR_EXPORT_URL=https://<hub-canister-id>.icp0.io

51

# KNOWTATION_OPERATOR_EXPORT_KEY=long-random-secret-matching-admin_set_operator_export_secret

52

# KNOWTATION_OPERATOR_EXPORT_DIR=./backups

53

#

54

# Optional: before `dfx deploy hub --network ic`, `npm run canister:release-prep` can export vault(s) to ./backups/

55

# (same script as `npm run canister:export-backup` and scheduled HTTP vault export — see docs/DEPLOY-HOSTED.md §6)

56

# KNOWTATION_CANISTER_BACKUP_USER_ID=google:123

57

# KNOWTATION_CANISTER_URL=https://<canister-id>.icp0.io

58

# KNOWTATION_CANISTER_BACKUP_URL= # alias for CANISTER_URL (e.g. CI secret name)

59

# KNOWTATION_CANISTER_BACKUP_VAULT_ID=default

60

# KNOWTATION_CANISTER_BACKUP_VAULT_IDS=default,second # comma-separated; overrides VAULT_ID when set

61

# KNOWTATION_CANISTER_BACKUP_DIR=./backups

62

# KNOWTATION_CANISTER_BACKUP_ENCRYPT_KEY_HEX= # 64 hex chars → AES-256-GCM .json.enc output

63

# KNOWTATION_CANISTER_BACKUP_S3_BUCKET=

64

# KNOWTATION_CANISTER_BACKUP_S3_PREFIX=knowtation-canister-backups/

65

# KNOWTATION_CANISTER_BACKUP_SKIP_S3=1 # local: skip S3 even if bucket set

66

# AWS_ACCESS_KEY_ID= AWS_SECRET_ACCESS_KEY= AWS_REGION=us-east-1

67

#

68

# CLI / curl / repo scripts (verify-hosted-hub-api, seed:*, `knowtation propose`) read KNOWTATION_HUB_URL,

69

# KNOWTATION_HUB_TOKEN, KNOWTATION_HUB_VAULT_ID — not *_LOCAL / *_HOSTED. If you use suffixed names in .env,

70

# after `source .env` run exports, e.g. for hosted:

71

# export KNOWTATION_HUB_URL="$KNOWTATION_HUB_URL_HOSTED"

72

# export KNOWTATION_HUB_TOKEN="$KNOWTATION_HUB_TOKEN_HOSTED"

73

# export KNOWTATION_HUB_VAULT_ID="$KNOWTATION_HUB_VAULT_ID_HOSTED"

74

#

75

# Operator human-lane smokes (laptop only — not Netlify/cron): one-time bootstrap:

76

# node scripts/hub-session-refresh.mjs --save-access-token '<hub_token from Local Storage>'

77

# Then `node scripts/hub-session-refresh.mjs` auto-renews (~/.config/knowtation/hub_refresh).

78

# Optional for legacy_session deploy probes only (copy from knowtation-gateway SESSION_SECRET):

79

# KNOWTATION_SESSION_SECRET=

80

# `node scripts/hub-session-refresh.mjs` rotates access JWT automatically (see script header).

81

# KNOWTATION_HUB_REFRESH_TOKEN_FILE=~/.config/knowtation/hub_refresh

82

# KNOWTATION_HUB_TOKEN_FILE=~/.config/knowtation/hub_session

83

#

84

# Optional proposal LLM (exact names; env locks Settings → Backup checkboxes when set to 1/0 or true/false):

85

# KNOWTATION_HUB_PROPOSAL_REVIEW_HINTS=1

86

# KNOWTATION_HUB_PROPOSAL_ENRICH=1

87

#

88

# AIR Improvement D: built-in attestation endpoint (hosted gateway).

89

# When set (32+ chars), the gateway auto-configures KNOWTATION_AIR_ENDPOINT to its own

90

# /api/v1/attest route. All hosted note writes are then attested (HMAC-signed, stored in

91

# Netlify Blobs). Verify any attestation via GET /api/v1/attest/:id.

92

# ATTESTATION_SECRET=your-long-random-secret-min-32-chars

93

#

94

# Hosted memory consolidation (Session 10 / Stream 1 — bridge + gateway).

95

# CONSOLIDATION_LLM_API_KEY: LLM key for hosted consolidation passes; falls back to OPENAI_API_KEY.

96

# CONSOLIDATION_LLM_MODEL: Model to use (default: gpt-4o-mini).

97

# CONSOLIDATION_COST_CAP_USD: Optional daily spend cap per user on the bridge (no cap if unset).

98

# CONSOLIDATION_LLM_API_KEY=

99

# CONSOLIDATION_LLM_MODEL=gpt-4o-mini

100

# CONSOLIDATION_COST_CAP_USD=

101

#

102

# Sign in with Apple — native identity-assertion exchange (POST /api/v1/auth/native-apple-exchange).

103

# Placeholders only — never commit real Bundle IDs, Team IDs, or .p8 material.

104

# APPLE_CLIENT_ID= # Required for Apple path: native Bundle ID or Services ID (JWT aud)

105

# APPLE_TEAM_ID= # Reserved (future auth-code / client-secret JWT) — do not commit

106

# APPLE_KEY_ID= # Reserved (.p8 key id) — do not commit

107

# APPLE_PRIVATE_KEY= # Reserved (.p8 PEM) — gitignored secret store only; never commit

108

#

109

# Docs connectors (KN-DOCS-SYNC) — names only; gates stay false in source until Tier 3.

110

# DOCS_OAUTH_GOOGLE_AUTHORIZED=

111

# DOCS_NOTION_HUB_KEY_AUTHORIZED=

112

# GOOGLE_DRIVE_OAUTH_CLIENT_ID=

113

# GOOGLE_DRIVE_OAUTH_CLIENT_SECRET=

114

# KNOWTATION_DOCS_OAUTH_SECRET=

115

# DOCS_OAUTH_REDIRECT_URI=

116

# SCOOLING_RETURN_URL_ALLOWLIST=

117

# NOTION_API_KEY=

118

#

119

# Learning-path writes (KN-WORK-PATH-LIST). Name only; default off until Tier 3.

120

# PATH_WRITES_ENABLED=