localagentbrowser 1.57__tar.gz → 1.62__tar.gz

{localagentbrowser-1.57 → localagentbrowser-1.62}/CHANGELOG.md

@@ -1,11 +1,69 @@
 # Changelog
 
-Every version of `agent_browser.py`, newest first. Each released version is also a tagged commit (`v1.11` … `v1.57`); 1.54 to 1.56 were internal steps
-measured on the regression suite and released as part of 1.57 (see
+Every version of LocalAgentBrowser, newest first. Each released version is also a tagged commit (`v1.11` … `v1.62`); 1.54 to 1.56 and 1.58 to 1.60 were internal steps
+measured on the regression suite and released as part of 1.57 and 1.61 (see
 [docs/regression-history.md](docs/regression-history.md)).
 Timings are from live runs on the development machine (Intel i5-9500, 6 cores, no GPU, Qwen2.5-7B-Instruct
 Q4_K_M in LM Studio), where reading the prompt takes about 90% of every model call.
 
+## [1.62] - 2026-10-25
+No change in what the agent does: the same prompts, decisions and results as 1.61.
+### Changed
+- The code is now a package, `localagentbrowser`, with one file per job (settings, prompts, page reading,
+  grounding, checklist, coverage, checks, report, and the loop) instead of one 1,805-line `agent_browser.py`. The
+  code was moved, not rewritten: every line of the old file is in the package once, in its original order with its
+  comments. `python agent_browser.py`, the `localagentbrowser` command and `import agent_browser` work as before.
+  The README's Code layout section says which file holds what.
+### Added
+- Golden prompts (`tests/test_golden_prompts.py`): five scripted runs, including one through the outline pointer,
+  a refusal of a finding not on the page, a coverage send-back with a fresh prompt and caveats, compared character
+  for character with a recorded copy of every message sent to the model, the journal and the report. It was
+  recorded before the move and passed unchanged after it; a one-word change to a check's wording makes it fail and
+  name the model call that differs.
+
+Regression run on the reorganized code: 7 of 7 finished (168.3 model minutes), against 6 of 7 (222.1) on 1.61. With
+the same behaviour this is run-to-run variation (the documentation task finished in 5 turns this time), not an
+improvement; each task's first prompt was identical to 1.61's, and every check still fired where its path came up.
+
+## [1.61] - 2026-10-21
+Released together with the internal steps 1.58, 1.59 and 1.60, each measured on the regression suite
+([docs/regression-history.md](docs/regression-history.md) has every step's results, task by task).
+### Added
+- An outline pointer for long pages. On a page long enough to have an outline, the read-on check looks for an unread
+  part whose label names a word of an item with nothing recorded yet ("Administration" for "setup and
+  administration"), and a plain READ_MORE then reads that part instead of the next one. Told only to use the
+  outline, the model went on to the next part 6 times out of 6 on a 10-part comparison page; with the code making
+  the jump, it reaches the part every time. Words that say what the whole page is about are ignored (the page title's
+  words, words every item shares), as are item verbs and parts that start inside code; hyphenated words stay whole.
+  On 1.61 the comparison finished for the first time since 1.56, from that page's Administration part.
+- "Every item has the discoveries it asks for: choose FINISH now" in the note after a discovery, when each item has
+  at least what it asks for. When that holds, the read-on check also asks for FINISH instead of offering to read on
+  (replayed: read on 6 of 6 before, FINISH 9 of 9 after).
+### Changed
+- The agent is sent back before finishing only for gaps the code verifies: an item nothing was copied for, fewer
+  things than the item asks for, or nothing recorded. The model's other verdicts are logged "(not sent back)" and
+  only shape the final result. About half of the send-backs for those verdicts had been wrong, costing finishes.
+- Every item that asks for a number is counted, not only items with a condition: an owl task finished with two
+  tribes for "at least three cultures". Amounts of time or measure ("the first two weeks") are not counts.
+- A thing not shown to meet a condition stays in the result, marked where it is first named ("**OnlyOffice** (not
+  shown in the pages read to be free and open-source)"), instead of being left out: leaving such things out dropped
+  correct answers along with wrong ones.
+- A description named by the coverage check as one of a list's things is ignored (it had made two office suites
+  count as three).
+- Smaller fresh prompts: refusal reasons in the journal are cut to 120 characters, pages list 8 links instead of 15
+  (those with a few words of text first; in four suites the agent opened 39 pages, none from page links), and links
+  whose text is HTML are skipped. Fresh prompts are about 2,500 tokens instead of 3,500-4,900.
+- Regression suite: the scorer accepts a forbidden item when the result marks it with the caveat.
+### Fixed
+- When several things shared a caveat, only the first was marked: the identical note next to it counted for all.
+
+Regression runs, finished out of 7 (model minutes): 1.58 5 (214), 1.59 5 (209.5), 1.60 6 (170.6), 1.61 6 (222.1).
+What is still open: a fresh prompt (when a conversation reaches its token budget) does not carry the advice to
+finish, and on 1.61 the documentation task read on after one with all four items found; nothing checks that a
+recipe's steps reach their end (on 1.60 cold brew stopped after step 2). Measured and not used: checking a
+discovery's link to its item by shared words (most links without a shared word are right), and a 16,384-token
+context, which would halve the fresh prompts but needs more memory than the default aims for.
+
 ## [1.57] - 2026-10-19
 Released together with the internal steps 1.54, 1.55 and 1.56, each of which was measured on the regression suite
 but not released on its own. [docs/regression-history.md](docs/regression-history.md) has every step's results,

{localagentbrowser-1.57 → localagentbrowser-1.62}/MANIFEST.in

@@ -2,3 +2,4 @@
 include CHANGELOG.md
 recursive-include tests *.py
 recursive-include tools *.py
+recursive-include tests *.json

{localagentbrowser-1.57/localagentbrowser.egg-info → localagentbrowser-1.62}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: localagentbrowser
-Version: 1.57
+Version: 1.62
 Summary: An autonomous web agent driven by a local LLM (LM Studio) and Playwright. Runs entirely on your own machine.
 Author: SnowFX
 License-Expression: MIT
@@ -199,19 +199,19 @@ does not generalise; the [changelog](https://github.com/SnowFXXX/LocalAgentBrows
 
 Since 1.53, every change is measured on the same seven objectives, one of each kind above, run one after the other
 with a time limit each (`tools/regression.py`, see [Testing](#testing)). Single runs vary with the pages a search
-turns up, so the suite compares versions on all seven at once. 1.54 to 1.56 were internal steps measured this way
-and released together as 1.57; [docs/regression-history.md](https://github.com/SnowFXXX/LocalAgentBrowser/blob/main/docs/regression-history.md)
+turns up, so the suite compares versions on all seven at once. 1.54 to 1.56 and 1.58 to 1.60 were internal steps
+measured this way and released with 1.57 and 1.61; [docs/regression-history.md](https://github.com/SnowFXXX/LocalAgentBrowser/blob/main/docs/regression-history.md)
 has every step's results and what each one showed.
 
-| Task | 1.52 | 1.53 (objective checklist) | 1.57 (coverage check) |
-|---|---|---|---|
-| Factual question | Finished, 12 min | Finished, 14 min | Finished, 13 min |
-| List of three office suites | Time limit | Finished, 30 min; but WPS Office is not open source | Time limit: three correct open-source suites by turn 7, sent back because no words showed them to be "free" |
-| Comparison and recommendation | Time limit | Finished, 39 min | Time limit, after a send-back for "official documentation", which the objective does not ask for |
-| How-to (cold brew from 1.55; repotting before) | Time limit | Time limit | Finished, 36 min; but "how long it keeps" answered with the steeping time |
-| Explanation (photosynthesis) | Time limit | Finished, 13 min, but thin: no oxygen or water | Finished, 59 min: inputs, outputs and both stages, cited |
-| Opinion (owls) | Time limit | Finished, 38 min | Time limit: its first page covered two of the three cultures asked for |
-| Documentation (venv, official docs) | Finished, 15 min, from a tutorial site | Time limit (30 min) | Finished, 37 min, all from python.org; but the Windows line repeats the create command instead of the activate one |
+| Task | 1.52 | 1.53 (objective checklist) | 1.57 (coverage check) | 1.61 (verified send-backs, outline pointer) |
+|---|---|---|---|---|
+| Factual question | Finished, 12 min | Finished, 14 min | Finished, 13 min | Finished, 17 min |
+| List of three office suites | Time limit | Finished, 30 min; but WPS Office is not open source | Time limit: three correct open-source suites by turn 7, sent back because no words showed them to be "free" | Finished, 21 min: LibreOffice, WPS Office and OnlyOffice, each marked "not shown in the pages read to be free and open-source" |
+| Comparison and recommendation | Time limit | Finished, 39 min | Time limit, after a send-back for "official documentation", which the objective does not ask for | Finished, 26 min, setup and administration from the part of a long page the outline pointer chose |
+| How-to (cold brew from 1.55; repotting before) | Time limit | Time limit | Finished, 36 min; but "how long it keeps" answered with the steeping time | Finished, 33 min: every step, the 16-hour steep and the shelf life, cited |
+| Explanation (photosynthesis) | Time limit | Finished, 13 min, but thin: no oxygen or water | Finished, 59 min: inputs, outputs and both stages, cited | Finished, 39 min |
+| Opinion (owls) | Time limit | Finished, 38 min | Time limit: its first page covered two of the three cultures asked for | Finished, 42 min: four cultures, cited, with a view |
+| Documentation (venv, official docs) | Finished, 15 min, from a tutorial site | Time limit (30 min) | Finished, 37 min, all from python.org; but the Windows line repeats the create command instead of the activate one | Time limit: all four items found from python.org, then it read on |
 
 ## How it works
 
@@ -237,6 +237,25 @@ flowchart LR
     C -->|FINISH| R[research_report.md]
 ```
 
+### Code layout
+
+The code is in the `localagentbrowser` package, one file per job, each depending only on the ones above it:
+
+| File | What it holds |
+|---|---|
+| `config.py` | Settings: the LM Studio connection (from environment variables) and every limit and threshold |
+| `prompts.py` | Every fixed text the model reads: the system prompt, turn instructions, the checklist, coverage and final-result prompts |
+| `text.py` | Small word helpers the checks share |
+| `pages.py` | Reading the web: search results, page parts, the outline and its pointer, links |
+| `grounding.py` | Whether a finding is really on the page shown, or already recorded |
+| `checklist.py` | The objective split into a fixed checklist, and findings linked to its items |
+| `coverage.py` | The coverage check before finishing, verified in code, and what is reported as not found or not proven |
+| `checks.py` | Decisions refused or confirmed before they are carried out (leaving, reading on, finishing, ...) |
+| `report.py` | The journal, the research log, the final result and the report file |
+| `agent.py` | The loop: each turn's conversation with the model, in-turn retries, and `main()` |
+
+`agent_browser.py` keeps `python agent_browser.py` working from a clone.
+
 ### Keeping a small model on track
 
 - **Rules enforced in code, not just stated in the prompt.** Repeated searches, visiting the current page,
@@ -307,8 +326,8 @@ the conversation nears the context limit. The journal uses a compact one-line-pe
 | `LMSTUDIO_API_KEY` | `lm-studio` | API key, if your server requires one |
 | `LMSTUDIO_CONTEXT_LENGTH` | `8192` | Context length the model is loaded with; the conversation budget follows it |
 
-Behaviour limits (part size, retry counts, check intervals and so on) are named constants at the top of
-`agent_browser.py`, each with a comment explaining it.
+Behaviour limits (part size, retry counts, check intervals and so on) are named constants in
+`localagentbrowser/config.py`, each with a comment explaining it.
 
 ## Performance
 
@@ -323,9 +342,9 @@ Measured on an Intel i5-9500 (6 cores, no GPU) with Qwen2.5-7B-Instruct Q4_K_M a
 | Coverage check before finishing (all checklist items) | about 1–3 min |
 | Writing the final result (once, at the end) | 50–190 s, depending on how much was found |
 
-Whole runs vary with the pages a search turns up. In the regression runs since 1.53, a factual question took 13 to
-19 minutes of model time; the list, comparison, how-to, opinion and documentation tasks 21 to 44 when they finished;
-and the explanation 13 to 59. A GPU or a smaller model will be much faster.
+Whole runs vary with the pages a search turns up. In the regression runs since 1.53, a factual question took 9 to
+19 minutes of model time; the list, comparison, how-to, opinion and documentation tasks 17 to 44 when they finished;
+and the explanation 13 to 59. A whole suite of seven tasks took 171 to 272 minutes. A GPU or a smaller model will be much faster.
 
 ## FAQ
 
@@ -363,15 +382,17 @@ turn that brings in a new page takes a few minutes. See [Performance](#performan
 - **Judging whether a finding covers an item is the weakest area.** The model's verdict is unreliable both ways:
   on 1.53 the explanation finished from one overview paragraph without oxygen or water, and in the regression runs
   since 1.54 about half of the coverage check's send-backs were for gaps that were not real ("missing official
-  documentation" on a comparison that did not ask for it, "H2O, O2" already recorded). The code now verifies what
-  it can (an item nothing answers, a count, word-for-word proof of a condition), but each wrong send-back still
-  costs a finish on this hardware, since a send-back costs 8–10 minutes.
+  documentation" on a comparison that did not ask for it, "H2O, O2" already recorded). Since 1.58 the agent is sent
+  back only for what the code verifies (an item nothing answers, fewer things than asked for); the model's other
+  verdicts only shape the final result, so a thin answer can still pass (on 1.61 the comparison's "many
+  simultaneous users" part is short).
 - **Conditions in the objective.** A small model tends to drop conditions such as "open source" or "using the
   official documentation": a list of open-source office suites once included FreeOffice (free but not open
   source), and a task that asked for the official Python documentation used a tutorial site. The objective
   checklist keeps the conditions in view (the documentation task uses python.org since 1.55), and a thing not
-  shown in the findings to meet a condition is left out of the result and listed after it. The check is strict:
-  on 1.57 it sent back three correct open-source suites because no words showed them to be "free". Check the
+  shown in the findings to meet a condition is marked where it is named ("not shown in the pages read to be free
+  and open-source"). The mark says what the pages read did not show, not that the thing fails the condition: on
+  1.61 it marked LibreOffice and OnlyOffice, which are open source, as well as WPS Office, which is not. Check the
   result against any conditions you set.
 - **A small model's judgement is the ceiling.** It sometimes drifts to loosely related pages, prefers
   secondary sources or stays with one source too long, and results vary with the first pages a search turns
@@ -402,8 +423,10 @@ turn that brings in a new page takes a few minutes. See [Performance](#performan
 - [x] Keep a fixed checklist of the objective's parts instead of re-splitting it at every check (1.53)
 - [ ] Check that a finding really covers its checklist item, including conditions like "open source" (partly, since
   1.57: a coverage check before finishing, with what the code can verify)
-- [ ] Send the agent back only for gaps the code verifies, and let the model's other verdicts only shape the final
-  result (the next step after 1.57)
+- [x] Send the agent back only for gaps the code verifies, and let the model's other verdicts only shape the final
+  result (1.58, released in 1.61)
+- [x] Point the agent to the part of a long page whose outline names an item it still needs (1.61)
+- [ ] Keep the advice to finish when a conversation starts over from a fresh prompt (the next step after 1.61)
 - [ ] Rewind dead ends: save the conversation before an expensive step such as reading a long page, and if it
   leads nowhere, return to that point with a short note instead of carrying the unused page text in every later
   prompt (LM Studio reuses its cache up to the saved point)
@@ -424,6 +447,12 @@ python -m pytest
 python -m ruff check .
 ```
 
+`tests/test_golden_prompts.py` replays five scripted runs and compares every message sent to the model, and the
+journal and report, with a recorded copy in `tests/golden/prompts.json`. A change meant to leave the agent's
+behaviour alone must keep it passing unchanged; a deliberate change to what the model sees is recorded again with
+`GOLDEN_UPDATE=1 python -m pytest tests/test_golden_prompts.py`, and the diff of the recorded copy reviewed with
+the rest of the change.
+
 To try the real agent with a fixed objective and a time limit, logging the time and tokens of every model
 call:
 
@@ -456,8 +485,8 @@ the [Code of Conduct](https://github.com/SnowFXXX/LocalAgentBrowser/blob/main/CO
 ## Project history
 
 Every released version is a tagged commit, and [CHANGELOG.md](https://github.com/SnowFXXX/LocalAgentBrowser/blob/main/CHANGELOG.md) describes what changed in each one and
-why, with the measurements behind the decisions. Versions 1.54 to 1.56 were internal steps, each measured on the
-regression suite and released together as 1.57; their results are in
+why, with the measurements behind the decisions. Versions 1.54 to 1.56 and 1.58 to 1.60 were internal steps, each
+measured on the regression suite and released with 1.57 and 1.61; their results are in
 [docs/regression-history.md](https://github.com/SnowFXXX/LocalAgentBrowser/blob/main/docs/regression-history.md).
 
 ## Acknowledgements

{localagentbrowser-1.57 → localagentbrowser-1.62}/README.md

@@ -167,19 +167,19 @@ does not generalise; the [changelog](https://github.com/SnowFXXX/LocalAgentBrows
 
 Since 1.53, every change is measured on the same seven objectives, one of each kind above, run one after the other
 with a time limit each (`tools/regression.py`, see [Testing](#testing)). Single runs vary with the pages a search
-turns up, so the suite compares versions on all seven at once. 1.54 to 1.56 were internal steps measured this way
-and released together as 1.57; [docs/regression-history.md](https://github.com/SnowFXXX/LocalAgentBrowser/blob/main/docs/regression-history.md)
+turns up, so the suite compares versions on all seven at once. 1.54 to 1.56 and 1.58 to 1.60 were internal steps
+measured this way and released with 1.57 and 1.61; [docs/regression-history.md](https://github.com/SnowFXXX/LocalAgentBrowser/blob/main/docs/regression-history.md)
 has every step's results and what each one showed.
 
-| Task | 1.52 | 1.53 (objective checklist) | 1.57 (coverage check) |
-|---|---|---|---|
-| Factual question | Finished, 12 min | Finished, 14 min | Finished, 13 min |
-| List of three office suites | Time limit | Finished, 30 min; but WPS Office is not open source | Time limit: three correct open-source suites by turn 7, sent back because no words showed them to be "free" |
-| Comparison and recommendation | Time limit | Finished, 39 min | Time limit, after a send-back for "official documentation", which the objective does not ask for |
-| How-to (cold brew from 1.55; repotting before) | Time limit | Time limit | Finished, 36 min; but "how long it keeps" answered with the steeping time |
-| Explanation (photosynthesis) | Time limit | Finished, 13 min, but thin: no oxygen or water | Finished, 59 min: inputs, outputs and both stages, cited |
-| Opinion (owls) | Time limit | Finished, 38 min | Time limit: its first page covered two of the three cultures asked for |
-| Documentation (venv, official docs) | Finished, 15 min, from a tutorial site | Time limit (30 min) | Finished, 37 min, all from python.org; but the Windows line repeats the create command instead of the activate one |
+| Task | 1.52 | 1.53 (objective checklist) | 1.57 (coverage check) | 1.61 (verified send-backs, outline pointer) |
+|---|---|---|---|---|
+| Factual question | Finished, 12 min | Finished, 14 min | Finished, 13 min | Finished, 17 min |
+| List of three office suites | Time limit | Finished, 30 min; but WPS Office is not open source | Time limit: three correct open-source suites by turn 7, sent back because no words showed them to be "free" | Finished, 21 min: LibreOffice, WPS Office and OnlyOffice, each marked "not shown in the pages read to be free and open-source" |
+| Comparison and recommendation | Time limit | Finished, 39 min | Time limit, after a send-back for "official documentation", which the objective does not ask for | Finished, 26 min, setup and administration from the part of a long page the outline pointer chose |
+| How-to (cold brew from 1.55; repotting before) | Time limit | Time limit | Finished, 36 min; but "how long it keeps" answered with the steeping time | Finished, 33 min: every step, the 16-hour steep and the shelf life, cited |
+| Explanation (photosynthesis) | Time limit | Finished, 13 min, but thin: no oxygen or water | Finished, 59 min: inputs, outputs and both stages, cited | Finished, 39 min |
+| Opinion (owls) | Time limit | Finished, 38 min | Time limit: its first page covered two of the three cultures asked for | Finished, 42 min: four cultures, cited, with a view |
+| Documentation (venv, official docs) | Finished, 15 min, from a tutorial site | Time limit (30 min) | Finished, 37 min, all from python.org; but the Windows line repeats the create command instead of the activate one | Time limit: all four items found from python.org, then it read on |
 
 ## How it works
 
@@ -205,6 +205,25 @@ flowchart LR
     C -->|FINISH| R[research_report.md]
 ```
 
+### Code layout
+
+The code is in the `localagentbrowser` package, one file per job, each depending only on the ones above it:
+
+| File | What it holds |
+|---|---|
+| `config.py` | Settings: the LM Studio connection (from environment variables) and every limit and threshold |
+| `prompts.py` | Every fixed text the model reads: the system prompt, turn instructions, the checklist, coverage and final-result prompts |
+| `text.py` | Small word helpers the checks share |
+| `pages.py` | Reading the web: search results, page parts, the outline and its pointer, links |
+| `grounding.py` | Whether a finding is really on the page shown, or already recorded |
+| `checklist.py` | The objective split into a fixed checklist, and findings linked to its items |
+| `coverage.py` | The coverage check before finishing, verified in code, and what is reported as not found or not proven |
+| `checks.py` | Decisions refused or confirmed before they are carried out (leaving, reading on, finishing, ...) |
+| `report.py` | The journal, the research log, the final result and the report file |
+| `agent.py` | The loop: each turn's conversation with the model, in-turn retries, and `main()` |
+
+`agent_browser.py` keeps `python agent_browser.py` working from a clone.
+
 ### Keeping a small model on track
 
 - **Rules enforced in code, not just stated in the prompt.** Repeated searches, visiting the current page,
@@ -275,8 +294,8 @@ the conversation nears the context limit. The journal uses a compact one-line-pe
 | `LMSTUDIO_API_KEY` | `lm-studio` | API key, if your server requires one |
 | `LMSTUDIO_CONTEXT_LENGTH` | `8192` | Context length the model is loaded with; the conversation budget follows it |
 
-Behaviour limits (part size, retry counts, check intervals and so on) are named constants at the top of
-`agent_browser.py`, each with a comment explaining it.
+Behaviour limits (part size, retry counts, check intervals and so on) are named constants in
+`localagentbrowser/config.py`, each with a comment explaining it.
 
 ## Performance
 
@@ -291,9 +310,9 @@ Measured on an Intel i5-9500 (6 cores, no GPU) with Qwen2.5-7B-Instruct Q4_K_M a
 | Coverage check before finishing (all checklist items) | about 1–3 min |
 | Writing the final result (once, at the end) | 50–190 s, depending on how much was found |
 
-Whole runs vary with the pages a search turns up. In the regression runs since 1.53, a factual question took 13 to
-19 minutes of model time; the list, comparison, how-to, opinion and documentation tasks 21 to 44 when they finished;
-and the explanation 13 to 59. A GPU or a smaller model will be much faster.
+Whole runs vary with the pages a search turns up. In the regression runs since 1.53, a factual question took 9 to
+19 minutes of model time; the list, comparison, how-to, opinion and documentation tasks 17 to 44 when they finished;
+and the explanation 13 to 59. A whole suite of seven tasks took 171 to 272 minutes. A GPU or a smaller model will be much faster.
 
 ## FAQ
 
@@ -331,15 +350,17 @@ turn that brings in a new page takes a few minutes. See [Performance](#performan
 - **Judging whether a finding covers an item is the weakest area.** The model's verdict is unreliable both ways:
   on 1.53 the explanation finished from one overview paragraph without oxygen or water, and in the regression runs
   since 1.54 about half of the coverage check's send-backs were for gaps that were not real ("missing official
-  documentation" on a comparison that did not ask for it, "H2O, O2" already recorded). The code now verifies what
-  it can (an item nothing answers, a count, word-for-word proof of a condition), but each wrong send-back still
-  costs a finish on this hardware, since a send-back costs 8–10 minutes.
+  documentation" on a comparison that did not ask for it, "H2O, O2" already recorded). Since 1.58 the agent is sent
+  back only for what the code verifies (an item nothing answers, fewer things than asked for); the model's other
+  verdicts only shape the final result, so a thin answer can still pass (on 1.61 the comparison's "many
+  simultaneous users" part is short).
 - **Conditions in the objective.** A small model tends to drop conditions such as "open source" or "using the
   official documentation": a list of open-source office suites once included FreeOffice (free but not open
   source), and a task that asked for the official Python documentation used a tutorial site. The objective
   checklist keeps the conditions in view (the documentation task uses python.org since 1.55), and a thing not
-  shown in the findings to meet a condition is left out of the result and listed after it. The check is strict:
-  on 1.57 it sent back three correct open-source suites because no words showed them to be "free". Check the
+  shown in the findings to meet a condition is marked where it is named ("not shown in the pages read to be free
+  and open-source"). The mark says what the pages read did not show, not that the thing fails the condition: on
+  1.61 it marked LibreOffice and OnlyOffice, which are open source, as well as WPS Office, which is not. Check the
   result against any conditions you set.
 - **A small model's judgement is the ceiling.** It sometimes drifts to loosely related pages, prefers
   secondary sources or stays with one source too long, and results vary with the first pages a search turns
@@ -370,8 +391,10 @@ turn that brings in a new page takes a few minutes. See [Performance](#performan
 - [x] Keep a fixed checklist of the objective's parts instead of re-splitting it at every check (1.53)
 - [ ] Check that a finding really covers its checklist item, including conditions like "open source" (partly, since
   1.57: a coverage check before finishing, with what the code can verify)
-- [ ] Send the agent back only for gaps the code verifies, and let the model's other verdicts only shape the final
-  result (the next step after 1.57)
+- [x] Send the agent back only for gaps the code verifies, and let the model's other verdicts only shape the final
+  result (1.58, released in 1.61)
+- [x] Point the agent to the part of a long page whose outline names an item it still needs (1.61)
+- [ ] Keep the advice to finish when a conversation starts over from a fresh prompt (the next step after 1.61)
 - [ ] Rewind dead ends: save the conversation before an expensive step such as reading a long page, and if it
   leads nowhere, return to that point with a short note instead of carrying the unused page text in every later
   prompt (LM Studio reuses its cache up to the saved point)
@@ -392,6 +415,12 @@ python -m pytest
 python -m ruff check .
 ```
 
+`tests/test_golden_prompts.py` replays five scripted runs and compares every message sent to the model, and the
+journal and report, with a recorded copy in `tests/golden/prompts.json`. A change meant to leave the agent's
+behaviour alone must keep it passing unchanged; a deliberate change to what the model sees is recorded again with
+`GOLDEN_UPDATE=1 python -m pytest tests/test_golden_prompts.py`, and the diff of the recorded copy reviewed with
+the rest of the change.
+
 To try the real agent with a fixed objective and a time limit, logging the time and tokens of every model
 call:
 
@@ -424,8 +453,8 @@ the [Code of Conduct](https://github.com/SnowFXXX/LocalAgentBrowser/blob/main/CO
 ## Project history
 
 Every released version is a tagged commit, and [CHANGELOG.md](https://github.com/SnowFXXX/LocalAgentBrowser/blob/main/CHANGELOG.md) describes what changed in each one and
-why, with the measurements behind the decisions. Versions 1.54 to 1.56 were internal steps, each measured on the
-regression suite and released together as 1.57; their results are in
+why, with the measurements behind the decisions. Versions 1.54 to 1.56 and 1.58 to 1.60 were internal steps, each
+measured on the regression suite and released with 1.57 and 1.61; their results are in
 [docs/regression-history.md](https://github.com/SnowFXXX/LocalAgentBrowser/blob/main/docs/regression-history.md).
 
 ## Acknowledgements

localagentbrowser-1.62/agent_browser.py

@@ -0,0 +1,23 @@
+"""LocalAgentBrowser: an autonomous web agent driven by a local language model.
+
+Run:  python agent_browser.py   (or `localagentbrowser` once installed; it asks for the objective)
+
+The code lives in the localagentbrowser package (see its __init__.py for how it is organized). This file keeps
+`python agent_browser.py` working from a clone, and `import agent_browser` working for the tests and tools: it
+gives access to every name in the package. To replace a part of the agent in a test, patch the module that
+uses it (for example localagentbrowser.agent.AsyncOpenAI), not this file.
+"""
+from localagentbrowser.config import *  # noqa: F403
+from localagentbrowser.prompts import *  # noqa: F403
+from localagentbrowser.text import *  # noqa: F403
+from localagentbrowser.pages import *  # noqa: F403
+from localagentbrowser.grounding import *  # noqa: F403
+from localagentbrowser.checklist import *  # noqa: F403
+from localagentbrowser.coverage import *  # noqa: F403
+from localagentbrowser.checks import *  # noqa: F403
+from localagentbrowser.report import *  # noqa: F403
+from localagentbrowser.agent import *  # noqa: F403
+from localagentbrowser.agent import main, run  # noqa: F401
+
+if __name__ == "__main__":
+    run()

localagentbrowser-1.62/localagentbrowser/__init__.py

@@ -0,0 +1,26 @@
+"""LocalAgentBrowser: an autonomous web agent driven by a local language model.
+
+A model served by LM Studio decides, one JSON action at a time, whether to SEARCH, NAVIGATE, READ_MORE,
+DISCOVER or FINISH; Playwright carries the actions out in a visible Chromium window. Everything the
+agent does and finds is written to research_log.txt.
+
+Run:  python agent_browser.py   (or `localagentbrowser` once installed; it asks for the objective)
+
+Settings (environment variables):
+    LMSTUDIO_BASE_URL        OpenAI-compatible server address   (default http://localhost:1234/v1)
+    LMSTUDIO_MODEL           model identifier                   (default qwen2.5-7b-instruct)
+    LMSTUDIO_API_KEY         API key, if the server needs one   (default lm-studio)
+    LMSTUDIO_CONTEXT_LENGTH  context length the model is loaded with (default 8192)
+
+The package, one file per job; each imports only from the ones listed before it:
+    config      settings and every limit and threshold
+    prompts     every fixed text the model reads
+    text        small word helpers the checks share
+    pages       search results, page parts, the outline and its pointer, links
+    grounding   whether a finding is really on the page shown, or already recorded
+    checklist   the objective split into a fixed checklist; findings linked to its items
+    coverage    the coverage check before finishing, verified in code
+    checks      decisions refused or confirmed before they are carried out
+    report      the journal, research log, final result and report file
+    agent       the loop: each turn's conversation with the model, and main()
+"""