*** BEST VIEWED WITH CURIOSITY ENABLED ***
Workflow plan
Pokémon TCG market podcast creation system
A complete production workflow for transforming Pokémon TCG news, product drops, market data, and PikaPikaPapa transcript context into a recurring podcast-style market briefing.
Core output
Audio
A polished market briefing script, rendered as MP3 and delivered through RSS.
Primary lens
TCG
Pokémon cards, sealed product, singles, grading, promos, and market behavior.
Context engine
Ryan
PikaPikaPapa transcripts and notes provide investment framework and historical takes.
Success rule
Live
The system reports success only after the published enclosure is reachable.
Mission
Build a repeatable market briefing, not a one-off summary.
The system should scan the Pokémon TCG market each day, separate meaningful signals from noise, connect those signals to PikaPikaPapa’s existing investing perspective, and produce a concise audio-first briefing that sounds like a knowledgeable host preparing collectors and investors for what to watch next.
The finished workflow should answer four questions every run: what happened, why it matters, how strong the signal is, and what should be watched next.
Episode shape
Default editorial structure.
- Market temperature open
- Top three market stories
- Product drop and restock watch
- Singles, sealed, and graded movement
- PikaPikaPapa context layer
- Quick hits and watchlist close
Source map
Each source has a purpose, extraction method, and reliability tier.
| Source | Purpose | Data captured | Use in show | Reliability |
|---|---|---|---|---|
| PikaPikaPapa transcripts and vault notes | Long-term investing lens | Set opinions, card theses, sealed logic, grading views, risk warnings | Context and “how this fits Ryan’s framework” segments | Primary context |
| PokéBeach | News and reveal tracking | Set reveals, Japanese product news, promos, card images, release timing | News selection, reveal analysis, quick hits | Primary news |
| Pokémon Center | Official product availability | Drops, restocks, sellouts, exclusives, bundles, ETBs, promos | Product watch and urgency scoring | Official |
| TCGPlayer Seller Blog | Marketplace and seller-side signals | Demand commentary, supply pressure, product movement, seller guidance | Market interpretation and trend confirmation | Market signal |
| TCGPlayer marketplace data | Price and liquidity movement | Market price, listed median, sales velocity proxies, sealed movement | Singles and sealed market segment | Quant signal |
| eBay sold listings | Real transaction comps | Recent sales, auction behavior, sealed premiums, slab comps | Price confirmation and outlier detection | Comp signal |
| PriceCharting / PSA / grading data | Graded market context | Raw vs graded spread, PSA 10 movement, population pressure | Grading economics and slab watchlist | Validate |
| Official Pokémon news | Release calendar and announcements | Official announcements, set dates, product descriptions, events | Source-of-truth confirmation | Official |
| Japanese retail and news sources | Early product signals | Japan releases, lottery demand, preorder behavior, specialty products | Forward-looking demand and English-market setup | Contextual |
Production gates
Every run moves through explicit contracts and artifacts.
Scheduled launchA timer starts the daily one-shot workflow, loads runtime configuration, creates a run ID, and opens a run folder.
Prerequisite checkVerify source credentials where needed, Python environment, feed parser, article extractor, LLM route, TTS engine, ffmpeg, RSS writer, and publisher access.
Source collectionFetch RSS feeds, official pages, product pages, seller blog posts, saved market data snapshots, and recent PikaPikaPapa transcript additions.
NormalizationConvert each item into a common story record: title, source, URL, timestamp, category guess, article text, product names, set names, card names, and confidence.
Repeat suppressionCompare against the story ledger so recurring product pages, repeated reveals, and syndicated summaries do not dominate multiple episodes.
Market scoringScore each item for market impact, novelty, urgency, investor relevance, collector hype, data confidence, and connection to known PikaPikaPapa themes.
Bucket selectionChoose deep dives, quick hits, rapid-fire notes, product watch items, data watch items, and skipped archive items using category caps.
Research packetsBuild evidence bundles for selected stories with primary article text, related coverage, market snapshots, transcript excerpts, and contradiction notes.
Script assemblyWrite the episode in audio-first sections, preserving source facts and separating confirmed data from interpretation.
Editorial reviewCheck for repetition, weak transitions, unsupported claims, missing caveats, awkward audio phrasing, and overhyped investment language.
Audio renderNormalize text for speech, synthesize audio, compress the MP3, write metadata, and store the final artifact in the run folder.
Publish and verifyUpdate RSS, publish files, fetch the live enclosure URL, validate the feed, then notify only after the live asset is reachable.
Story scoring model
The score decides airtime.
- Market impact
- How likely the item is to affect sealed, singles, grading, or product demand.
- Novelty
- Whether this is new information or a repeat of known news.
- Urgency
- Whether collectors need to act, watch, or ignore today.
- Ryan alignment
- How strongly the item connects to PikaPikaPapa themes and prior commentary.
- Evidence strength
- Whether the story has official confirmation, market data, or only sentiment.
- Audience value
- Whether the story teaches something useful beyond “this happened.”
Default airtime thresholds
- 85–100: lead story or deep dive.
- 70–84: quick hit with context.
- 50–69: rapid-fire mention or watchlist.
- Below 50: archive only.
Writing architecture
Separate judgment, reporting, voice, and editing.
| Role | Input | Output | Quality rule |
|---|---|---|---|
| Classifier | Normalized source records | Category, entities, confidence, initial score | Must identify product, set, card, and market bucket when present. |
| Market analyst | Scored records plus price/product snapshots | Market implication notes and risk level | Must explain why the signal matters or mark it as noise. |
| PikaPikaPapa context retriever | Selected story entities and themes | Relevant transcript excerpts and historical commentary | Must quote or summarize only retrieved context, not invent prior positions. |
| Research packet builder | Article text, official info, market data, transcript context | Evidence packet with source facts, interpretation angles, and uncertainty | Must separate confirmed facts from market read. |
| Host writer | Episode outline and research packets | Conversational spoken script | Must sound natural out loud and avoid screen-first formatting. |
| Editor | Draft script | Clean final script with tighter pacing | Must remove repetition, hype inflation, unsupported certainty, and abrupt transitions. |
| Metadata writer | Final script and manifest | Episode title, summary, chapters, RSS description | Must be concise and accurate to the finished audio. |
Audio-first rules
The final script is meant to be heard.
- Use short spoken paragraphs and clear transitions.
- Read prices, dates, set names, and acronyms in a TTS-friendly way.
- Say “this is a watch item” when data is early or incomplete.
- Never imply guaranteed investment outcomes.
- Prefer useful framing over hype: supply, demand, timing, entry point, liquidity, and risk.
- End each deep dive with a practical “what to watch next” note.
Run artifacts
Files make the system debuggable.
config/sources.json
config/scoring.json
config/runtime.json
var/runs/<run_id>/manifest.json
var/runs/<run_id>/source-items.json
var/runs/<run_id>/selected-stories.json
var/runs/<run_id>/research-packets/
var/runs/<run_id>/script-final.txt
var/runs/<run_id>/audio-final.mp3
var/story-ledger.json
var/product-ledger.json
var/price-snapshots/
rss/pokemon-market-briefing.xml
Audio, RSS, and publication workflow
A successful episode is one that can be played from the published feed.
Speech normalizationConvert symbols, prices, percentages, Japanese set names, card abbreviations, and dates into speech-safe text.
TTS renderingRender the final script with the preferred voice and speed, then create a fallback render path for failed synthesis.
Audio processingTrim silence, normalize loudness, compress to MP3, and embed title, date, episode ID, and artwork metadata.
Feed updateAppend a new RSS item with stable GUID, title, summary, publication date, duration, byte size, and enclosure URL.
PublishUpload or commit the MP3 and RSS feed to the hosting target.
VerifyFetch the public RSS feed, parse the latest item, request the enclosure URL, confirm status, content type, and nonzero size.
NotifySend the completion summary with episode title, top stories, duration, and verified audio URL.
Failure semantics
Optional misses are acceptable. Broken episodes are not.
| Type | Examples | Behavior |
|---|---|---|
| Fatal | No selected stories, LLM unavailable, script missing, audio missing, RSS invalid, published enclosure unreachable | Stop the run, preserve artifacts, retry once, then report the failed gate. |
| Recoverable | One feed fails, one article extraction fails, one price source is unavailable, a transcript match is weak | Continue with lower confidence and record the missing source in the manifest. |
| Editorial | Too much repeated coverage, weak source diversity, overlong draft, unsupported market claim | Rewrite the affected section before audio rendering. |
| Success | Final audio exists, RSS validates, live enclosure responds, manifest is complete | Mark the run complete and send the verified episode summary. |
Build roadmap
Recommended order of construction.
Phase 1: Source inventoryDefine source list, fetch cadence, source priority, extraction method, and terms for each Pokémon TCG market feed.
Phase 2: PikaPikaPapa retrieval layerIndex transcripts and notes by set, card, product type, era, grading topic, sealed topic, and investment theme.
Phase 3: Written briefing MVPGenerate a daily text briefing with scored stories, source links, Ryan-context notes, and a watchlist.
Phase 4: Scoring and ledgersAdd repeat suppression, product ledger, scoring thresholds, category caps, confidence labels, and per-run manifests.
Phase 5: Research packet systemCreate structured packets for lead stories that combine news, official info, market data, and PikaPikaPapa context.
Phase 6: Audio script writerTransform packets into a host-quality spoken script with intro, deep dives, quick hits, watchlist, and close.
Phase 7: Editorial gateReview and rewrite for accuracy, pacing, transitions, claim discipline, and audio clarity.
Phase 8: TTS renderNormalize speech text, render MP3, process audio, store artifacts, and collect duration/size metadata.
Phase 9: RSS publishingGenerate the podcast feed, publish audio, validate feed structure, and verify the live enclosure.
Phase 10: Automation and monitoringSchedule the workflow, add retry behavior, notify completion, summarize failures, and retain run history for debugging.