Repository navigation
pricing: add Agent37 - #557
vishnukool wants to merge 2 commits into
Conversation
Contributor License AgreementAll contributors are covered by a CLA. |
![devin-ai-integration[bot]](https://avatars.githubusercontent.com/in/811515?s=60&v=4){kind=small}
There was a problem hiding this comment.
Devin Review found 2 potential issues.
1 flag not posted on this PR by your GitHub settings — view it in Devin Review. (Configure)
| "total_1vcpu_2gb_hr": 0.003014, | ||
| "confidence": "exact", | ||
| "notes": "($0.80 + 2 x $0.70) / 730 hours. Smallest shape is 2 vCPU / 4 GB. Performance instances, whose cores are held for them, are 4x on compute." |
There was a problem hiding this comment.
🔴 Agent37's displayed cost halves its minimum
For Agent37, total_1vcpu_2gb_hr prices a shape its smallest instance cannot provide. getEffectiveCost displays $0.0030/hour, although the minimum 2-vCPU, 4-GB instance costs $0.0060/hour before disk.
Learn more
The pricing chart uses getEffectiveCost to sort and show each provider's hourly cost, falling back to normalized.total_1vcpu_2gb_hr for wall-clock providers getEffectiveCost. Agent37's notes say its minimum is 2 vCPU and 4 GB, so the advertised 1-vCPU, 2-GB combination cannot be purchased. The chart will show half the compute charge for a minimum-sized instance, before disk charges.
Example: A customer runs the smallest Agent37 instance for one hour. CPU and RAM cost (2 × $0.80 + 4 × $0.70) / 730 ≈ $0.00603, but the comparison shows $0.0030.
Recommended fix: Keep the theoretical normalization if it is needed for comparisons, but supply effective_cost_1vcpu_2gb_hr for Agent37 at the minimum billable shape, or make the chart explicitly compare hypothetical shapes rather than actual purchasable hourly costs. Regenerate the chart after updating the data.
Was this helpful? React with 👍 or 👎 to provide feedback.
There was a problem hiding this comment.
Agreed. Added effective_cost_1vcpu_2gb_hr: 0.006027 (the 2 vCPU / 4 GB shape), so the chart shows $0.0060.
| { | ||
| "id": "agent37", |
There was a problem hiding this comment.
🟡 Public pricing chart omits Agent37
Adding Agent37 only to pricing.json leaves the README's embedded pricing.svg unchanged. The pricing workflow runs only on manual dispatch, so the published comparison continues to omit Agent37 until regeneration.
Learn more
The README embeds the committed chart directly pricing comparison. Its generator reads the JSON and writes the SVG, but the pricing workflow does not run on changes to the JSON. A weekly benchmark can eventually regenerate it, but the PR itself ships an unchanged comparison.
Example: Immediately after merging this PR, a README visitor sees the old chart with no Agent37 row, even though the data file lists Agent37.
Recommended fix: Regenerate and commit the SVG with the new provider, or add an automatic regeneration workflow triggered by pricing data changes.
Was this helpful? React with 👍 or 👎 to provide feedback.
There was a problem hiding this comment.
Regenerated pricing.svg in this PR; the only change is the new Agent37 row.
Adds Agent37 to
pricing.json, with rates from https://www.agent37.com/pricing: $0.80 per vCPU, $0.70 per GB of RAM and $0.09 per GB of disk per 730-hour month, metered by the minute. That makes 1 vCPU + 2 GB $0.0030 an hour, and our smallest shape (2 vCPU, 4 GB) $0.0060 an hour, which is the cost the chart shows. The chart is regenerated with the new row.The provider package is in computesdk/computesdk#1013.
benchmarkstays pending until Agent37 is in a TTI run.