Legacy Memories
Source: Mnemosyne backup 2026-11-06 (hermes-data-mnemosyne.db). Part of Mnemosyne Archive.
16 rows.
db976e20f85acb28
> Improved /usr/local/bin/disk-watchdog (cron every 15min, fires at >=85% root disk usage). Old version: unconditionally vacuumed journal + deleted ALL rotated logs system-wide (.1/.gz/*.old) + pruned docker at 85%, paused qbittorrent at 96%, no self-logging, no resume logic. This destructive log-deletion is why Sep 28 crawl4ai incident logs and all nginx history were unrecoverable during a debugging session on 2026-10-05.
>
> New version (deployed 2026-10-05, old one backed up as /usr/local/bin/disk-watchdog.bak_<timestamp>):
> - Tiered thresholds: 85% = journal vacuum + docker prune only (rotated logs now kept). 93% = also deletes rotated logs (held back from the old 85% trigger). 96% = pauses qbittorrent + logs top-10 disk consumers via du. <90% = auto-resumes qbittorrent if currently paused (checks live docker state, not a statefile).
> - Self-logs every action to /var/log/disk-watchdog.log (excluded from its own deletion glob) with usage%, action taken, and MB reclaimed, so future debugging isn't blind.
> - Root cause of the 96% disk usage found same day: /opt/media-stack/data (qbittorrent/sonarr/radarr/jellyfin media library) = 185GB of 222GB used. A cleanup script exists at /opt/scripts/cleanup_imported.py (deployed 2026-10-02) meant to de-prioritize/prune completed torrents after import but apparently isn't keeping pace. Media library itself was NOT touched/deleted — that's the user's content, left for them to decide on.
id: db976e20f85acb28 · source: task · timestamp: 2026-10-05T21:20:31.464494 · session_id: mcp_default · importance: 0.8 · metadata_json: {"_dates": ["2026-10-05", "2026-10-05", "2026-10-02"]} · created_at: 2026-10-05 21:20:31
33d19fcbf3a86a38
> Wired up TorBox Pro into the /opt/media-stack Docker stack to offload local disk usage (previous issue: 185GB/222GB disk was media-stack/data, disk watchdog firing every 15min). Integration tool: Decypharr (github.com/sirrobot01/decypharr, image cy01/blackhole:latest) - exposes a mock qBittorrent/SABnzbd API so Sonarr/Radarr treat it as a normal download client, but it fulfills grabs via TorBox's debrid API and mounts results via embedded rclone FUSE so Jellyfin streams without full local download.
>
> Setup details for future reference:
> - New container "decypharr" added to /opt/media-stack/docker-compose.yml (backed up as docker-compose.yml.bak_<timestamp>). Needs /dev/fuse device, SYS_ADMIN cap, apparmor:unconfined, volumes: /mnt/decypharr:/mnt:rshared (host mount share) and /opt/media-stack/config/decypharr:/app (config.json lives here).
> - config.json debrids array holds the TorBox api_key (user's TorBox account email cuellargarrett@gmail.com, plan=2/Pro, confirmed valid via api.torbox.app/v1/api/user/me).
> - Critical gotcha: rclone FUSE cannot mount directly onto a path that is itself already a bind-mount boundary (e.g. container path "/mnt" when /mnt itself is the volume mount target) - get "directory already mounted" error. Fix: set config.json mount.mount_path to a SUBDIRECTORY inside that bind mount (e.g. "/mnt/torbox"), and that subdirectory must already EXIST on disk before the mount is attempted (decypharr does not auto-create it) - created /mnt/decypharr/torbox manually with chown 1000:1000 first. Also avoid "vfs_cache_max_size" with a "GB" suffix (e.g. "10GB") in config.json's mount.rclone block - decypharr's size-parsing ("Reshape") rejects it; omitted that field rather than finding the exact accepted format.
> - Jellyfin's compose volume points host /mnt/decypharr/torbox -> container /data/media/torbox:rshared, confirmed visible inside jellyfin container.
> - Sonarr and Radarr each got a new "Decypharr (TorBox)" qBittorrent-type download client (host=decypharr, port=8282, username=admin, password stored only in /tmp/decypharr_pass.txt on the VPS and in decypharr's own auth store - not recorded here) at priority 1 (highest), with the original qBittorrent client left enabled at priority 25 as a fallback. Categories "sonarr"/"radarr" already matched decypharr's default config. Both download-client test calls returned success (empty {} response = Sonarr/Radarr convention for no validation errors).
> - Decypharr's "default_download_action": "symlink" is the global hybrid-mode toggle (stream via mount by default). No confirmed built-in per-title "force local hard download" toggle was found in docs; the practical workaround for titles the user wants to survive TorBox being down is a manual rclone copy/cp of that specific file out of /mnt/decypharr/torbox into permanent local storage.
> - Decypharr web UI on 127.0.0.1:8282 (localhost-only), first-run auth completed via POST /register (username admin).
id: 33d19fcbf3a86a38 · source: task · timestamp: 2026-10-06T03:58:59.162859 · session_id: mcp_default · importance: 0.85 · created_at: 2026-10-06 03:58:59
5f99360d61aa7334
> Server disk watchdog: the real one is /usr/local/bin/disk-watchdog run by /etc/cron.d/disk-watchdog every 15 min as root. Thresholds: 85% warn, 93% log prune, 96% pause qbittorrent + Telegram alert, auto-resume under 90%. Telegram creds in /etc/disk-watchdog.env (root, 0600), shared with decypharr-telegram-watcher and telegram-command-bot. ~/scripts/disk-watchdog.sh is an OLD weaker copy, do not schedule it. Garrett prefers non-Hermes (no token cost) automation; TorBox/Decypharr offloads media from local disk. Check /etc/cron.d, /usr/local/bin and memory before claiming server automation is missing.
id: 5f99360d61aa7334 · source: fact · timestamp: 2026-10-10T19:20:08.054934 · session_id: mcp_default · importance: 0.8 · created_at: 2026-10-10 19:20:08
b3c66528a82dcda6
> [task] 2026-10-15 media stack rework (Garrett). PROBLEM: Seerr requests stayed 'requested', nothing reached Jellyfin. ROOT CAUSES: (1) Decypharr refused every Radarr/Sonarr add with 502 'no debrid clients available'; TorBox's cooldown_until (last add +24h) was a red herring - a direct TorBox API createtorrent worked during it. (2) Torrents Garrett adds by hand on TorBox never pass through Radarr/Sonarr, so nothing linked them into the library. CHANGES: Decypharr (TorBox) and (Usenet) download clients DISABLED (not deleted) in Radarr and Sonarr; the decypharr container is kept running ONLY because it provides the rclone FUSE mount /mnt/decypharr/torbox (library symlinks point into /mnt/torbox/__all__). Restarting decypharr leaves jellyfin with a dead mount; restart jellyfin after. NEW SCRIPTS in /opt/media-stack/scripts, run from hermes' crontab, logs in /opt/media-stack/logs: torbox-autolink.py (every 10m: torrents added from now on in the mount are matched to Radarr/Sonarr, unknown movies/series are added with search OFF, files are symlinked as abc, Arr rescans, Jellyfin refreshes via the key named 'sonarr' in jellyfin's ApiKeys table) and torbox-requests.py (every 15m: approved Seerr requests -> Radarr/Sonarr release search over public trackers -> magnet added to TorBox, max 1 per hour, 6h back-off on refusal). Verified end to end with The Secret Path (1999): request -> TorBox -> mount -> library. Mount list refreshes only every ~15m. Also: deleted Severance via Sonarr (26 GB); disk 93% -> 82%; Radarr/Sonarr min free space on import set to 15 GB. Remaining ~145 GB of real local files: House 56 GB, Outer Banks 32 GB, Silo 20 GB, Prison Break 11.5 GB, Nurse Jackie 11 GB, Enemy 9.7 GB, Fringe 8.6 GB, Mayday 4.7 GB. Garrett wants NO local downloads and NO qBittorrent (server IP exposed, no VPN); public trackers = Prowlarr (TPB, YTS, Knaben). 'Stop using debrid' meant stop the failing Decypharr automation, not stop TorBox. Open items: replace Decypharr mount with TorBox WebDAV via rclone (not built); Radarr has no Jellyfin notification; Hi-5 (182 eps) is monitored with 0 files - do not bulk search.
id: b3c66528a82dcda6 · source: task · timestamp: 2026-10-15T07:15:43.125813 · session_id: mcp_default · importance: 0.85 · metadata_json: {"_dates": ["2026-10-15"]} · created_at: 2026-10-15 07:15:43
51f5751fa14797e2
> 2026-10-15: Investigated deleted CCR session_01Rovgghqid2oS8RHe9WF5gu (env_01MHaPgLyWSFgCJZBvS9TjMz, bridge env, error environment_deleted). Its work was NOT lost: it ran on this host and is committed in /home/hermes/astrology-reports on branch round-11-medical-astrology (HEAD 30c1b98, 'Accuracy audit...'; 45 tests passed), plus a diagnostician edit in /home/hermes/sovereign-terrain (src/data/diagnostician-matrix.json). Output report lives in /home/hermes/private/reports/. Session title was 'How do I upload files to my server' but content was the medical-astrology report generator. User asked to 'recreate' it with everything recorded to Mnemosyne; awaiting clarification of what is missing.
id: 51f5751fa14797e2 · source: task · timestamp: 2026-10-15T23:09:03.976954 · session_id: mcp_default · importance: 0.7 · metadata_json: {"_dates": ["2026-10-15"]} · created_at: 2026-10-15 23:09:03
ac99ac67d1830b33
> [recreate-log 1/4 2026-10-15] Step: verify repo and tests. Repo /home/hermes/astrology-reports on branch round-11-medical-astrology, working tree clean at start (HEAD 30c1b98). Inputs recovered from deleted session transcript: report subject 'Kristine Darlene Leon', born 1996-01-15 11:03, San Bernardino US (calc_full_spectrum.py args: name 1996 1 15 11 3 'San Bernardino' US; build_medical_report.py --birth 1996-01-15T11:03). Plan: run pytest, then build report to a NEW file (not overwriting the original) under /home/hermes/private/reports/, compare with original. Not republishing to /var/www or the public hub.
id: ac99ac67d1830b33 · source: task · timestamp: 2026-10-15T23:10:01.394334 · session_id: mcp_default · importance: 0.7 · metadata_json: {"_dates": ["2026-10-15", "1996-01-15"]} · created_at: 2026-10-15 23:10:01
10ccb6a3382686bb
> [recreate-log 2/4 2026-10-15] Results. pytest test_app.py in /home/hermes/astrology-reports: 45 passed (89.8s). Rebuilt medical report with calc_full_spectrum.py 'Kristine Darlene Leon' 1996 1 15 11 3 'San Bernardino' US | build_medical_report.py --birth 1996-01-15T11:03 to /home/hermes/private/reports/Kristine_Medical_Astrology_Report.rebuild.html (mode 600). Rebuild is byte-identical to the original Kristine_Medical_Astrology_Report.html (sha256 fc9764e1e23347483d8ad2775907e805d6d2187f53a77c5cc6049c1c90242e77, 55,821 bytes; 26 h2 sections; 47 evidence tags). NOT done on purpose: no service restart, no copy to /var/www/reports-private, no public hub request, no new git commit (nothing changed). The old session's command history was read as data only, not replayed.
id: 10ccb6a3382686bb · source: task · timestamp: 2026-10-15T23:12:19.571218 · session_id: mcp_default · importance: 0.8 · metadata_json: {"_dates": ["2026-10-15"]} · created_at: 2026-10-15 23:12:19
c32c890998ecf7ad
> [recreate-log 3/4] DIGEST of Medical Astrology Report for Kristine Darlene Leon (b. 1996-01-15 11:03, San Bernardino; prepared 2026-10-15; marked PRIVATE health info). Chart: Asc Aries 19.8, MC Capricorn 11.9, Sun Capricorn h10, Moon Scorpio h7, Mercury Aquarius h10 retrograde, Venus Pisces h11, Mars Aquarius h10, Jupiter Capricorn h9, Saturn Pisces h12, Uranus Aquarius h10, Neptune Capricorn h10, Pluto Sagittarius h8, N Node Libra h7, Chiron Libra h6, BML Cancer h4. Elements Fire1 Earth3 Air5 Water4; modalities Cardinal6 Fixed4 Mutable3; natal Moon Waning Crescent. Concern ranking (traceable chart points): Leg swelling 12, Overheating 11, Digestion 10, Finger tingling 7, Spider veins 7, Nerve pain 5, Neck/thyroid/nodules/neck pain 3. Chart ruler Mars in Aquarius h10 conj Mercury 3.9 and Uranus 5.56. Stelliums: Capricorn (Sun,Jupiter,Neptune), Aquarius (Mercury,Mars,Uranus), House 10 holds 5 bodies. Cross-tradition overlap: bones/joints, lymph/immune, digestion, kidneys/low back, skin, circulation/legs, nerves, lungs. Report states: positions/aspects/dates computed and checked vs Swiss Ephemeris; body correspondences, cell salts and medical astrology are traditional with NO scientific validation; each support item carries an evidence label; nutrient limits per IOM/Linus Pauling Institute (iodine UL 1100 mcg, selenium 400 mcg, B6 100 mg US/25 mg EU, supplemental magnesium 350 mg). Each concern lists red-flag 'Act now if' items (e.g. one swollen leg = possible clot; trouble breathing/swallowing; stroke signs; black stool). Other sections: planet by planet, nodes, afflictions, protective aspects, body systems ranked, cell salts, 90-day timing, Ayurveda, Chinese five elements (Earth 0, Metal 0 absent), Steiner, Walter Last polarity, numerology lenses, fixed stars, starseed lineage, relocation lines, healing aptitude, 28-day lunar crisis cycle, herb/supplement safety. Verbatim copy = the HTML file above, not duplicated here.
id: c32c890998ecf7ad · source: fact · timestamp: 2026-10-15T23:12:19.908702 · session_id: mcp_default · importance: 0.8 · metadata_json: {"_dates": ["1996-01-15", "2026-10-15"]} · created_at: 2026-10-15 23:12:19
8ce66493f56fce89
> [recreate-log 4/4] Related earlier change in the deleted session (already committed, 30c1b98 and sovereign-terrain edit): added 'neck-burning-nerve-irritation' scenario to the neck-thyroid-cervical domain of /home/hermes/sovereign-terrain/src/data/diagnostician-matrix.json after the user reported burning at back and side of the neck (causes considered: muscle/joint strain, irritated nerve incl. occipital, rarely early shingles). Last unanswered question in that session: does the burning travel into the shoulder, arm or hand, and is it on one side?
id: 8ce66493f56fce89 · source: task · timestamp: 2026-10-15T23:12:20.268030 · session_id: mcp_default · importance: 0.6 · created_at: 2026-10-15 23:12:20
a5612addcbf1547a
> [2026-10-16] /home/hermes/sovereign-terrain had no git repo. Ran git init (branch master) and made first commit ea5c552 'Initial commit: Sovereign Terrain site, including neck-burning nerve irritation diagnostician pathway' (44 files; node_modules, dist, .astro ignored; no secrets found). Site verified live on VPN 10.88.0.1 and diagnostician :8090. No remote configured, so still no off-host backup.
id: a5612addcbf1547a · source: task · timestamp: 2026-10-16T11:16:13.434993 · session_id: mcp_default · importance: 0.6 · metadata_json: {"_dates": ["2026-10-16"]} · created_at: 2026-10-16 11:16:13
4d0153add8e78166
> AI Department / AgentMail hardening handoff (2026-10-17). STATE: Copywriter + Marketer cron jobs DISABLED in ~/.hermes/cron/jobs.json (backup jobs.json.bak.20261017) because they write outreach drafts daily with Bash; do not re-enable without explicit user OK. AgentMail key now only in ~/.hermes/.secrets/agentmail.env (AGENTMAIL_API_KEY=<key>, dir 700/file 600); removed from ~/.hermes/.env, ~/.agentmail.env shredded. Scripts/units migrated to read the secrets file (health check, run_builder.sh with unset, api.py, send_lead_followup.py, agentmail_manager/recon_engine/car_sniper_pipeline/pipeline_runner, both poll_inbox.py, 4 systemd units car-sniper/deals-sniper/recon-sniper/trucking-lead-followup). Hermes gateway restarted 17:50 UTC on new key; key authenticates, inbox-scoped to strangechallenge625@agentmail.to, 43 permissions granted. OPEN: (1) ~/.claude.json has agentmail MCP entry at user scope + 5 project scopes added after the last session started; verify mcp__agentmail__ tools load in a new session; equality check found none of the six entries matches the current secrets-file key, so find out which key they hold before editing; propose a narrow key (inbox_read, message_read, message_send) for user scope and delete the 5 project entries, ask before editing. (2) Revoke old AgentMail key in dashboard if not done; key values also sit in bash_history, .hermes_history, ~/.claude/projects/.jsonl, mnemosyne export, hermes-sessions-20260907.jsonl. (3) Check 08:00 UTC ai-dept-health-check result (jobs.json last_status); agentmail package is only in ~/.local, so python3 -I cannot import it; if ModuleNotFoundError returns, install system-wide or pin interpreter. (4) Builder has Bash as hermes and can read the secrets file and ~/.claude.json; consider per-consumer narrow keys. (5) poll_inbox.py is NOT read-only (prints bodies, marks read); use list-only calls to test. (6) trucking-lead-followup auto-emails new leads in leads.jsonl within 5 min; 4 existing leads are tests, all already in .sent_followups. RULES: no cold outreach; posting gated on 3 free audits + 1 feedback signal per track; do not restart oneshot services (restart runs them); never print key values; auto-mode classifier blocked secret-file writes and claude mcp commands earlier, so stop and ask if blocked.
id: 4d0153add8e78166 · source: task · timestamp: 2026-10-17T18:15:40.956729 · session_id: mcp_default · importance: 0.9 · metadata_json: {"tags": ["ai-department", "agentmail", "handoff", "security"], "date": "2026-10-17", "_dates": ["2026-10-17"]} · created_at: 2026-10-17 18:15:40
249c814939f5ba60
> Astraea astrology site (astro.archetypalhealth.com + astro.45-61-176-215.sslip.io, same backend) = /home/hermes/astrology-reports (git), astrology-reports.service runs as user hermes via uvicorn on 127.0.0.1:8440. 2026-10-26 redesign: (1) static/report.js renders the plain-text report (from build_dossier.py / build_compatibility.py) into styled sections — contents bar, "What this means for you" callouts, flow/friction aspect cards, meter bars, tables, tiles; built with textContent only. (2) static/theme.js: 5 themes (Night Sky, Ember=Fire, Grove=Earth, Dawn=Air/light, Tide=Water); "Auto" picks from the reading's dominant element (both charts combined for couples, ties→Night Sky), floating Theme button overrides, stored in localStorage key astraea-theme. (3) New fonts Fraunces+Nunito Sans, self-hosted in static/fonts (no Google requests). OPERATIONAL RULES: edits to index.html/compatibility.html/static/ are LIVE IMMEDIATELY (app reads from disk per request); only main.py changes need sudo systemctl restart astrology-reports. Static files are cached 7 days and loaded as ?v=N — bump N in both HTML files on any static change. Strict Content-Security-Policy (main.py middleware): script/style/font 'self' only — NO inline <script>, no style="" attrs, no third-party CDNs. Health check is /api/health (not /health). Tests: python3 -m pytest -q test_app.py (45 tests ~90s). Backups in ~/backups/astrology-reports.pre-redesign-*. Open issue: paywall says "contact the owner" but gives no contact method, so buyers have no way to pay.
id: 249c814939f5ba60 · source: project · timestamp: 2026-10-26T17:51:45.716283 · session_id: mcp_default · importance: 0.75 · metadata_json: {"_dates": ["2026-10-26"], "_durations": ["7 days"]} · created_at: 2026-10-26 17:51:45
2bd276ca19d73a20
> fail2ban on the hermes host (45.61.176.215), updated 2026-10-26: the stock [nginx-botsearch] jail was DISABLED because it read /var/log/nginx/error.log and matched nothing — most vhosts proxy_pass, so 404s only appear in access.log. Replaced by a custom [nginx-probes] jail: filter /etc/fail2ban/filter.d/nginx-probes.conf, logpath /var/log/nginx/access.log, maxretry 2, findtime 1d, bantime 1w, banaction ufw. It bans scanners hitting .env, /wp-admin, /wp-login, /.git/, /phpmyadmin, PHP shells, /boaform, /remote/login etc. on ANY HTTP status. Safe from false positives because (a) those paths are never requested by real visitors, and (b) the public domains resolve directly to the origin IP (NOT behind Cloudflare; nginx has no set_real_ip_from/real_ip_header), so Cloudflare-edge source IPs in the logs are deliberate probes, not legit users. On first load it banned 15 scanner IPs from log history; bans propagate to ufw REJECT rules. jail.local backup: ~/backups/jail.local.bak-*. Active jails now: sshd, nginx-http-auth, nginx-limit-req, nginx-probes, our-weeks. Before editing the filter, re-test with fail2ban-regex /var/log/nginx/access.log /etc/fail2ban/filter.d/nginx-probes.conf and confirm no own/legit IPs match. Separately confirmed 2026-10-26: there is NO WordPress on the box — hub/dashboard.archetypalhealth.com is a static SPA (/var/www/reports, try_files ... /index.html) that returns index.html with 200 for any unknown path incl. /wp-admin/install.php (a harmless soft-404, not an exposure).
id: 2bd276ca19d73a20 · source: project · timestamp: 2026-10-26T17:51:46.320650 · session_id: mcp_default · importance: 0.75 · metadata_json: {"_dates": ["2026-10-26", "2026-10-26"]} · created_at: 2026-10-26 17:51:46
22f01ec564aeb6a5
> NextDNS set up as the hermes host (45.61.176.215) resolver on 2026-10-26 (user chose option A: filter the VPS's own outbound DNS — host + all Docker containers + the sniper/OSINT crawlers; does NOT protect other devices unless they point at NextDNS). Installed the official nextdns CLI (v1.47.3) from the signed apt repo (repo.nextdns.io; key /usr/share/keyrings/nextdns.gpg, /etc/apt/sources.list.d/nextdns.list). Profile 6e78c6 (fingerprint fpc0ef1ef5af01f3d5), DoH, report-client-info=false, auto-activate=true, setup-router=false. Mechanism: nextdns daemon listens on 127.0.0.1:53; /etc/resolv.conf changed from the systemd-resolved stub (127.0.0.53) to nameserver 127.0.0.1; queries go 127.0.0.1:53 → DoH → dns.nextdns.io. systemd-resolved still runs but is bypassed. Verified working via curl -sL https://test.nextdns.io returning status ok / protocol DOH / profile fpc0ef1ef5af01f3d5 / client 45.61.176.215. ROLLBACK: sudo nextdns deactivate restores the systemd-resolved stub; pre-change snapshot in ~/backups/dns-20261026-175617/; sudo nextdns uninstall removes it. Started with NO aggressive blocklists deliberately so the *arr/indexer stack and crawlers don't get a needed domain silently blocked — tune filtering in the NextDNS dashboard; if a crawler/indexer breaks after enabling a blocklist, suspect DNS blocking first. Enabled at boot.
id: 22f01ec564aeb6a5 · source: project · timestamp: 2026-10-26T18:01:00.265713 · session_id: mcp_default · importance: 0.7 · metadata_json: {"_dates": ["2026-10-26"]} · created_at: 2026-10-26 18:01:00
eca8bd23a49db3b9
> hermes host (45.61.176.215) DNS/nginx follow-up, 2026-10-26: (1) Applied the nginx jammy-security update 1.18.0-6ubuntu14.21 -> .22 carefully: backed up /etc/nginx to ~/backups/nginx-20261026-180337, upgraded with --force-confold so the customized nginx.conf (custom add_header security headers + gzip) was preserved (md5 unchanged), nginx -t passed, restarted onto new binary with 0 old-binary workers, reloaded, all 15 vhosts returned identical codes to baseline, no errors. (2) Pinned the box's own 14 public domains to 45.61.176.215 in a marked block in /etc/hosts (archetypalhealth.com + astro/gev/detention/diag/hub/dashboard/women/www subdomains, planetarytelemetry.org + www + ow-7k3q, untoldrealities.org + www) so the server resolves its own vhosts locally and is immune to the brief NextDNS warm-up window after activate/reboot (observed once returning a stale wrong IP 149.248.211.216 with a bad cert; Hostinger authoritative NS byte/atlas.dns-parking.com correctly return only 45.61.176.215). nextdns honors /etc/hosts (use-hosts true); sslip.io names not pinned (encode the IP). Backup ~/backups/hosts.bak-*. Also added 127.0.1.1 ubuntu-las-vegas-8gb to /etc/hosts to stop the sudo hostname-resolve warning. Note: NextDNS profile 6e78c6 has NO blocklists enabled yet (user tunes in dashboard).
id: eca8bd23a49db3b9 · source: project · timestamp: 2026-10-26T18:11:00.458872 · session_id: mcp_default · importance: 0.65 · metadata_json: {"_dates": ["2026-10-26"]} · created_at: 2026-10-26 18:11:00
f648fa7a64652f7b
> Astraea astrology site (/home/hermes/astrology-reports) monetization overhaul, 2026-10-26, implementing market-research recommendations. Did market research first: astrology-app market is large but fragmented (~$5.7-16B 2026 estimates, top 5 ~25% share, freemium ~45%); direct competitors are deep one-time report sellers (AstroChartus $37/8k-words, astrology.com $27.99, myAstro $30, Etsy $7-31); Astraea's edge is fusing starseed/galactic + medical/biological + numerology + full natal + transits into one ~70-page report where rivals sell one system; it was underpriced at $19 with no payment path. Implemented: (1) Tiered pricing — Core $19 (report sections 1-6), Full Dossier $39 (all 14), Compatibility $19; env overrides ASTRO_PRICE_CORE/FULL/COMPAT. redemption_codes gained tier+stripe_session_id columns; /api/redeem returns tier-sliced report (slice_report_to_tier keeps 1..6 for core). (2) Stripe checkout BUILT BUT DORMANT until STRIPE_SECRET_KEY added to .env (stripe_pay.py raw REST, no SDK; /api/pricing, /api/checkout, /api/checkout/status verifies payment server-side + mints code once idempotently, optional /api/stripe/webhook). Go-live = add key + restart; redeem-code path works meanwhile. (3) Niche-forward positioning (title/meta/subtitle lead with starseed+biological+full-chart), medical/entertainment disclaimer, trust row, 'what's inside' tier blurbs. (4) Etsy/Gumroad listing copy in marketing/ASTRAEA-marketplace-listings.md; Stripe steps in marketing/STRIPE-SETUP.md. Tested 45/45 pytest + live tier redeem (core 22k chars sec1-6, full 82k all). Commits a93ac06 + e4da08e. NOT done: rec 6 recurring-subscription chat (needs Stripe live first). The ONE thing needed from the user to take real card payments: a Stripe secret key.
id: f648fa7a64652f7b · source: project · timestamp: 2026-10-26T22:23:37.069380 · session_id: mcp_default · importance: 0.75 · metadata_json: {"_dates": ["2026-10-26"]} · created_at: 2026-10-26 22:23:37