Travenza Holidays · Built by aayush · September 2026
Summary
The team is generating 100+ decks a month on the tool. The follow-up call set the priorities: day-wise content is the deal breaker, the cost reduction is confirmed, and images can wait for a second phase. This proposal delivers exactly that — ₹40,000, two months (August–September), broken up per workstream.
Four workstreams, each priced so scope stays visible: day-wise content quality (the deal breaker), accounts, history & follow-up edits, cost optimization, and ownership handoff. The body-font fix the team asked for is already in place, so it is not part of this engagement. Images move to a follow-on phase.
₹40,000
fixed total · 50% upfront · per-workstream pricing
70–85%
target reduction in per-deck model cost
Aug–Sep
2 months · month 2 is cover + handoff
What the call changed
Why decks take 40–50 minutes today
Generation itself works — the team confirmed the plane info, train info, and pricing pages hold up on real decks. The time goes into manual rework after generation: day-wise text gets rewritten in ChatGPT, and around 80% of images get swapped. That is where the 15-minute promise broke, and it is what this engagement is built to remove.
The before/after decks the team shared make the gap concrete. Generated day slides are short bullet lists — activity, times, one line each. The shipped versions are 150–250 words of flowing narrative per day: context, timings in the title, a closing line. That rewrite is where the time goes, and it is the acceptance target for the content workstream.
The team's own list of what would take the tool to 90% efficiency:
Team's 90% mark Verbatim from the call
"Better and more relevant images, itinerary history and prompts after pasting our excel, better detailed descriptions for day wise activities. The font size of the body is 20 right now and it's spoiling the alignment completely. Can we make this 16? If these things work, our tool will be at 90% efficiency."
The font fix is already handled. Three of the remaining items are in this phase; images move to a follow-on phase — they are their own build (a selection rework, not a tweak), and they land on top of the architecture this phase puts in place.
In this phase
Better detailed descriptions for day-wise activities — narrative quality matching the team's final decks, no ChatGPT rewrite.
Itinerary history and saved prompts after pasting the Excel — with per-user accounts and follow-up edit instructions.
The cost reduction update — confirmed on the call.
Deferred to the follow-on phase
Image relevance rework (quoted separately once this phase's selection architecture is live).
Desktop app — held; all work here is web-first on the existing browser app.
The Engagement
Four workstreams, priced separately
Nothing outside the list below is implied. Each workstream carries its own price so scope stays visible and bounded.
A
Workstream A · Day-wise content quality
Day slides the team can send without rewriting
The deal breaker from the call. Day-wise slides currently read as short bullet lists, so the team rewrites them into full narrative in ChatGPT before sending. This workstream makes the generated copy the copy they keep.
Rebuild day-wise generation to produce narrative prose per day — context, activity flow, timings in the title, a closing line — matching the richness of the final decks.
Retune the underlying prompts and model routing for this step specifically.
Acceptance: on a fixed sample, generated day-wise slides match the narrative quality of the before/after decks the team provided — no text rework needed.
₹14,000
B
Workstream B · Accounts, history & follow-up edits
Every user their own workspace, like ChatGPT
Today a pasted spreadsheet becomes a deck and the input is lost. The operations team asked directly: "Is there a way we can keep track of itineraries made and see a history?" This workstream makes the tool a workspace, not a pipeline.
Per-user accounts: every user signs in to their own workspace (auth already exists in the web app — this completes it).
Projects & history: each pasted itinerary is saved with its Excel input, prompt, and generated deck — revisitable and re-generatable.
Follow-up instructions: after a deck is generated, the user can send instructions to edit it — the tool re-generates with the changes applied.
₹10,000
C
Workstream C · Cost optimization
Lower per-deck cost, models switchable per node
Confirmed on the call: the team definitely wants the cost reduction update. Today each successful deck costs roughly ₹55 in model spend. This workstream restructures that.
OpenRouter routing: all generation moves to one integration with per-node model selection — models switchable from a config table.
Quality gate: cutover only after a fixed sample of itineraries matches today's quality bar.
September cover: cost-per-deck and failure-rate monitoring, with retuning if a model underperforms on real decks.
₹10,000
D
Workstream D · Ownership handoff
The product becomes an owned asset
Transfer operational control of the product to Travenza's own accounts — priced as its own workstream so the boundary is explicit.
Repositories under a Travenza-controlled GitHub (or equivalent) organisation.
Cloud Run / GCP project, secrets, and API billing under Travenza.
Deploy notes and a runbook so your team, or any future vendor, can ship without aayush.
₹6,000
Out of scope
Image relevance rework — the follow-on phase, quoted separately.
Desktop app work — held; this phase is web-first.
New PPT templates, slide redesigns, or new product features.
ERP integration changes or new data-source connectors.
Multi-tenant SaaS packaging or white-label sales channels.
What it does
Day-wise slides generated as narrative prose — rich enough to ship, no ChatGPT rewrite step.
Every user gets their own account, saved projects, and history of pasted itineraries.
Follow-up instructions after generation — edit and re-generate in the tool, not outside it.
All generation routed through OpenRouter with per-node model selection, targeting 70–85% cost reduction.
Repositories, GCP project, secrets, and billing transferred to Travenza accounts with a runbook.
Build Timeline
August ships, September covers
All four workstreams delivered within August. September is cutover follow-through, monitoring, and any cleanup.
Content quality + accounts & history
August · Week 1–2
Rebuild day-wise copy generation to narrative quality; per-user accounts, saved projects, and the follow-up edit flow.
OpenRouter integration + pilot
August · Week 2–3
Wire OpenRouter, port the node calls, run the quality-gate pilot on the fixed sample set against today's decks.
Cutover + ownership handoff
August · Week 3–4
Quality gate pass → cutover; transfer repositories, GCP project, secrets, and billing to Travenza accounts; write the runbook.
Operating cover
September
Watch cost per generation and failure rate; fix migration issues; retune any step that underperforms on real decks.
Cost
Fixed fee, broken up, ₹40,000 total
A fixed-fee engagement. Each workstream is priced so scope is visible; nothing is time-and-materials while models are being swapped.
Workstream breakdown
Workstream
Price
≈ USD
A · Day-wise content quality
₹14,000
$147
B · Accounts, history & follow-up edits
₹10,000
$105
C · Cost optimization
₹10,000
$105
D · Ownership handoff
₹6,000
$63
Total
₹40,000
$421
USD at ≈ ₹95/USD.
Payment
AugustA + B delivered · C integration and pilot · D handoff₹20,000
SeptemberOperating cover · monitoring and retuning₹20,000
Payment50% upfront (₹20,000) to start · ₹20,000 at the beginning of SeptemberPrepaid per month
After SeptemberOptional month-to-month cover at ₹12,000, or stopOptional
Total · 2 months₹40,000
Per successful deck
Model cost todayMeasured production baseline~₹55 (~$0.58)
Target after OpenRouter switchPilot confirms the final figure~₹8–17 (~$0.08–0.18)
Reduction70–85%
Monthly API spend at volume (illustrative)
Decks / month
Current (≈)
Target (≈)
Monthly savings (≈)
50
₹2,750
₹500
₹2,250
100
₹5,500
₹1,000
₹4,500
150
₹8,250
₹1,500
₹6,750
200
₹11,000
₹2,000
₹9,000
Assumptions: ~₹55 model cost per success today; target ~₹10 at the middle of the band. At the team's stated 100+ decks/month, the API saving alone is ~₹4,500/month — and it stays after the engagement ends.
Business Outcome
What changes for Travenza
Day-wise slides ship as generated — no ChatGPT rewrite step on the team's critical slides.
Every pasted itinerary is saved with its prompt and deck — history, re-generation, and follow-up edits in the tool.
Every user works in their own account — no shared logins, no lost work.
Per-deck model cost drops 70–85% (~₹4,500/month saved at 100 decks) — permanently.
Travenza owns the repositories, GCP project, and billing — an asset, not a rental.
A fixed, broken-up fee replaces open-ended invoices: ₹40,000, two payments, September as cover.
aayush built the current pipeline (an eleven-node LangGraph system that reads a raw itinerary and produces a complete, branded PPT deck) and runs it in production on Cloud Run. The stabilization engagement delivered ~30 changes across API and frontend — image reliability, generation hardening, layout updates, and platform security — verified end to end with seven test decks. The web app already carries login, deck storage, and an image-edit screen; this phase completes what is already half-built.
Conclusion
Content first. Cost down. ₹40,000.
The call set the priorities: day-wise content is the deal breaker, the cost reduction is confirmed, images follow later. This proposal delivers the content half of the team's 90% mark plus the cost restructuring, in two months (August–September), at ₹40,000 — broken up per workstream, prepaid per month. The image phase and the desktop app stay on the shelf until this lands.