Store reviewer logs as markdown files

[skaltman]

Change trajectory reviewer log format from .jsonl to Markdown.

Each conversation gets its own Markdown file. Each Markdown file has a YAML header and a Markdown body:

• The YAML front matter contains the flags, notes, and metadata needed to restore review state in the reviewer app.
• The Markdown body contains the conversation and repeats its flags and notes for the purposes of human readability. The YAML remains the source of truth for the flags and notes.

Locally, the review files are written to a directory set by the COMMONS_REVIEW_DIR environment variable or the review_dir arg to trajectory_review(). On Connect, they are stored in a pin so they persist across sessions.

Markdown format

---
generated_by: commons::trajectory_review
schema_version: 1
conversation: 20260728t222656.055-r0ai9xw7ic
last_active: 2026-07-28T15:33:48-0700
updated_at: '2026-08-14T22:14:01Z'
flags:
  conversation: no
  exchanges:
    - 1
notes:
  - user: unknown
    time: '2026-08-14T21:58:01Z'
    text: note about question 1
    exchange: 1
  - user: unknown
    time: '2026-08-14T22:13:45Z'
    text: note on question with searched context
    exchange: 3
  - user: unknown
    time: '2026-08-14T22:14:01Z'
    text: note on question with search pool
    exchange: 1
---

# 20260728t222656.055-r0ai9xw7ic

<!-- Generated by commons::trajectory_review(). Manual edits are not supported. -->

## Conversation-level review

**Flagged:** No

**Review notes:** None

## Exchange 1

**Trust:** Verified (A)

**Flagged:** Yes

**Review notes:**

> note about question 1
>
> _unknown, 2026-08-14T21:58:01Z_

> note on question with search pool
>
> _unknown, 2026-08-14T22:14:01Z_

### User

> How many patients had at least one serious adverse event, by arm?

### Tool call: `search_pool`

#### Arguments

 ```json
  {
    "query": "patients with at least one serious adverse event by arm"
  }
```
### Tool result: `search_pool`

_Discovery result omitted from the review log._

### Tool call: `call_measure`

#### Arguments

```json
{
  "name": "ae_overview",
  "arguments": "{\"population\": \"SAF\"}"
}
```

### Tool result: `call_measure`

|statistic                                                  |A: Drug X   |B: Placebo |C: Combination |
|:----------------------------------------------------------|:-----------|:----------|:--------------|
|Total number of patients with at least one AE              |100 (74.6%) |98 (73.1%) |103 (78.0%)    |
|Total number of AEs                                        |502         |480        |604            |
|Total number of deaths                                     |25 (18.7%)  |23 (17.2%) |22 (16.7%)     |
|Total number of patients withdrawn from study due to an AE |3 (2.2%)    |6 (4.5%)   |5 (3.8%)       |
|Total number of patients with at least one                 |NA          |NA         |NA             |
|Serious AE                                                 |85 (63.4%)  |80 (59.7%) |87 (65.9%)     |
|Related Serious AE                                         |64 (47.8%)  |52 (38.8%) |64 (48.5%)     |
|Related AE                                                 |86 (64.2%)  |85 (63.4%) |92 (69.7%)     |
|Severe AE (at greatest intensity)                          |77 (57.5%)  |70 (52.2%) |79 (59.8%)     |
|Grade 3-5 AE                                               |90 (67.2%)  |83 (61.9%) |93 (70.5%)     |
|Grade 4-5 AE                                               |77 (57.5%)  |70 (52.2%) |79 (59.8%)     |

Result stored as `r1`, available in your R session.
A data frame with 11 rows and 4 columns:
* statistic: character with 0 NAs, and 11 unique values
* A: Drug X: character with 1 NAs, and 9 unique values ("100 (74.6%)", "502", "25 (18.7%)", "3 (2.2%)", "85 (63.4%)", "64 (47.8%)", "86 (64.2%)", "77 (57.5%)", "90 (67.2%)")
* B: Placebo: character with 1 NAs, and 9 unique values ("98 (73.1%)", "480", "23 (17.2%)", "6 (4.5%)", "80 (59.7%)", "52 (38.8%)", "85 (63.4%)", "70 (52.2%)", "83 (61.9%)")
* C: Combination: character with 1 NAs, and 9 unique values ("103 (78.0%)", "604", "22 (16.7%)", "5 (3.8%)", "87 (65.9%)", "64 (48.5%)", "92 (69.7%)", "79 (59.8%)", "93 (70.5%)")

### Assistant

> Using the ae_overview measure (safety population), patients with at least one serious adverse event by arm:
>
> - Arm A (Drug X): 85 (63.4%)
> - Arm B (Placebo): 80 (59.7%)
> - Arm C (Combination): 87 (65.9%)

Rendered

--- generated_by: commons::trajectory_review schema_version: 1 conversation: 20260728t222656.055-r0ai9xw7ic last_active: 2026-07-28T15:33:48-0700 updated_at: '2026-08-14T22:14:01Z' flags: conversation: no exchanges: - 1 notes: - user: unknown time: '2026-08-14T21:58:01Z' text: note about question 1 exchange: 1 - user: unknown time: '2026-08-14T22:13:45Z' text: note on question with searched context exchange: 3 - user: unknown time: '2026-08-14T22:14:01Z' text: note on question with search pool exchange: 1 ---

20260728t222656.055-r0ai9xw7ic

Conversation-level review

Flagged: No

Review notes: None

Exchange 1

Trust: Verified (A)

Flagged: Yes

Review notes:

note about question 1

unknown, 2026-08-14T21:58:01Z

note on question with search pool

unknown, 2026-08-14T22:14:01Z

User

How many patients had at least one serious adverse event, by arm?

Tool call: search_pool

Arguments

 {
   "query": "patients with at least one serious adverse event by arm"
 }

Tool result: search_pool

Discovery result omitted from the review log.

Tool call: call_measure

Arguments

{
  "name": "ae_overview",
  "arguments": "{\"population\": \"SAF\"}"
}

Tool result: call_measure

statistic	A: Drug X	B: Placebo	C: Combination
Total number of patients with at least one AE	100 (74.6%)	98 (73.1%)	103 (78.0%)
Total number of AEs	502	480	604
Total number of deaths	25 (18.7%)	23 (17.2%)	22 (16.7%)
Total number of patients withdrawn from study due to an AE	3 (2.2%)	6 (4.5%)	5 (3.8%)
Total number of patients with at least one	NA	NA	NA
Serious AE	85 (63.4%)	80 (59.7%)	87 (65.9%)
Related Serious AE	64 (47.8%)	52 (38.8%)	64 (48.5%)
Related AE	86 (64.2%)	85 (63.4%)	92 (69.7%)
Severe AE (at greatest intensity)	77 (57.5%)	70 (52.2%)	79 (59.8%)
Grade 3-5 AE	90 (67.2%)	83 (61.9%)	93 (70.5%)
Grade 4-5 AE	77 (57.5%)	70 (52.2%)	79 (59.8%)

Result stored as r1, available in your R session.
A data frame with 11 rows and 4 columns:

• statistic: character with 0 NAs, and 11 unique values
• A: Drug X: character with 1 NAs, and 9 unique values ("100 (74.6%)", "502", "25 (18.7%)", "3 (2.2%)", "85 (63.4%)", "64 (47.8%)", "86 (64.2%)", "77 (57.5%)", "90 (67.2%)")
• B: Placebo: character with 1 NAs, and 9 unique values ("98 (73.1%)", "480", "23 (17.2%)", "6 (4.5%)", "80 (59.7%)", "52 (38.8%)", "85 (63.4%)", "70 (52.2%)", "83 (61.9%)")
• C: Combination: character with 1 NAs, and 9 unique values ("103 (78.0%)", "604", "22 (16.7%)", "5 (3.8%)", "87 (65.9%)", "64 (48.5%)", "92 (69.7%)", "79 (59.8%)", "93 (70.5%)")

Assistant

Using the ae_overview measure (safety population), patients with at least one serious adverse event by arm:

• Arm A (Drug X): 85 (63.4%)
• Arm B (Placebo): 80 (59.7%)
• Arm C (Combination): 87 (65.9%)

[github-actions]

Preview deployed to Connect (dogfood.team.pct.posit.it): https://dogfood.team.pct.posit.it/connect/#/apps/d7a36cae-8f27-448b-a478-61b81fbe3942/draft/361570

Deployed from commit a6b457d.

[github-actions]

Preview deployed to Connect (connect.staging.pct.posit.it): https://connect.staging.pct.posit.it/connect/#/apps/ad662e1b-5048-4acc-9ad7-f9478c92274e/draft/1804

Deployed from commit a6b457d.