You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
@@ -27,8 +27,7 @@ This project exists because Docling and Marker are both good tools, but in pract
27
27
28
28
- Async OCR requests and batch PDF processing
29
29
- Concurrent figure downloads per PDF
30
-
- Safe overwrite behavior with explicit `--overwrite`
31
-
- Progress display on `stderr` (TTY only)
30
+
- Fast! It takes approximately 25 sec per batch. The speed depends on the z.ai API availability. See
32
31
33
32
## Usage
34
33
@@ -90,6 +89,6 @@ Options:
90
89
91
90
`paperdown` reads `ZAI_API_KEY` from environment first. If not found, it reads the value from `--env-file`.
92
91
93
-
## Cost example
92
+
## Cost
94
93
95
-
The tool records token usage in `log.jsonl` under the `usage` field. With pricing at `$0.03` per `1,000,000` tokens, the Batista et al. (2022) paper run with `total_tokens = 79,080` costs `$0.0023724`. That is roughly `0.24` cents per paper.
94
+
The tool records token usage in `log.jsonl` under the `usage` field. With pricing at `$0.03` per `1,000,000` tokens, the average size scientific paper ([Batista et al., 2022](https://hess.copernicus.org/articles/26/3753/2022/)) process with `total_tokens = 79,080` costs `$0.0023724`. That is roughly `0.24` cents per article.
0 commit comments